Final Report Summary - PREDICT (Predicting individual response and resistance to VEGFR/mTOR pathway therapeutic intervention using biomarkers discovered through tumour functional genomics)
Executive Summary:
The main goal of the PREDICT consortium was to implement a research strategy that could identify predictive biomarkers of response to novel therapeutic interventions in renal cancer. Ours was a comprehensive bed-to-bench-to-bed research strategy that started with clinical samples, analyzed using the most powerful molecular profiling technologies including next generation sequencing. Additional biomarker leads were generated by siRNA based knock-down technologies, which were validated in newly established kidney cancer cell lines as well as isogenic cell lines. Finally, all these data were queried in the validation cohort of clinical samples.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Project Context and Objectives:
Effective translational research requires the integration of 1) understanding and clearly formulating the clinical problem at hand; 2) procurement and molecular characterization of appropriate clinical samples; 3) identification and application of the most relevant experimental models systems to the clinical problem; 4) translating the basic research results back into actual clinical practice.
The Objectives of the PREDICT Consortium followed the above described research strategy:
Objective 1: Understanding and clearly formulating the clinical problem at hand. Considering the large average size of primary renal tumours and our awareness of the existing scientific indications of intratumor heterogeneity we had to explore the issue of intratumor heterogeneity in a scientifically rigorous, quantifiable manner.
Objective 2: procurement and molecular characterization of appropriate clinical samples
Objective 2A: procurement of surgical material of primary renal tumours prior to everolimus or TKI treatment
Objective 2B: Sequencing of tumour samples
Objective 2C: Considering the importance of tyrosine kinase activity in the mechanism of effective TKI or everolimus treatment and the known discrepancy between the genomic and kinase activity changes, we designed experiments to directly measure kinase activity in the surgical specimens.
Objective 3: identification and application of the most relevant experimental models systems to the clinical problem
Objective 3A: Considering the limited number of cell lines for renal cancer cell lines available research, we embarked upon establishing novel RCC cell lines.
Objective 3B: Considering the limitations of cell line based experiments to the actual human clinical samples we aimed to establish a valuable panel of xenograft model systems for the renal cancer research community.
Objective 3C: Considering the limitations of associative analysis in biomarker research we established powerful causative discovery strategies by implementing several siRNA screen based research strategies.
Project Results:
Patient samples (WORK PACKAGE 1): One of the most important ingredients of translational research is the establishment of a carefully curated, well documented biobank of appropriately selected patient material. This was achieved to the highest possible standard permitted by changing clinical practice. Our original plans were significantly influenced by the fact that during the course of the PREDICT project the clinical use of mTOR inhibitors, everolimus has significantly changed. Since more effective first line treatments are available for RCC, everolimus had become a 2nd or 3rd line treatment option. This had made it both impossible and unethical to recruit the number of patients originally planned to be treated with everolimus. Yet, prior to this shift in the therapeutic use everolimus, in the context of E-PREDICT and NEORAD clinical trials we recruited 32 patients with careful clinical annotation and comprehensive outcome and follow-up data. We collected multiregionally sampled nephrectomy specimens from all 7 E-PREDICT cases and in two of those cases we also collected matched metastatic tissue. These tumor samples served and will serve at least two important purposes. First, these cases formed the basis of two mile-stone publications that effectively established the field of genomic analysis of intratumor heterogeneity (Gerlinger et al., 2014; 2012). These works have been covered and discussed by several editorials, reviews and we will mention here only the most important findings. While it has been known for a long time that tumor cells show significant heterogeneity the resolution of ITH at single nucleotide resolution was lacking. We applied multiregion whole exome sequencing to resolve the genetic architecture and evolutionary histories of ten ccRCCs. Ultra-deep sequencing identified ITH in all cases. We found that 73-75% of identified ccRCC driver aberrations were subclonal, confounding estimates of driver mutation prevalence. ITH increased with the number of biopsies analyzed, without evidence of saturation in most tumours. Chromosome 3p loss and VHL aberrations were the only ubiquitous events. Multi-regional tumour sampling was also performed on the cases enrolled in the NEORAD trial. We obtained sets for 23 cases consisting of: pre-treatment biopsy, 3 regions from the primary tumour and matched normal. The analysis of these samples is ongoing (see below for time estimates).
PREINSUT: We have 14 cases consisting of pre-treatment and post-treatment biopsy and matched normal tissue. 7 cases could not be included either due to poor quality of the isolated DNA or lack of matched normal tissue. Without the latter genomic data are extremely challenging to interpret, and therefore we proceeded with profiling of 7 cases in total.
SuMR: Of the 14 cases for which tissue was available (consisting of pre-treatment and post-treatment biopsy and matched normal tissue) 13 cases were suitable to sequencing.
SURTIME: We have received 7 cases with a single biopsy of the primary and matched normal. We are expecting to receive material for further 6 cases by June 2015.
PANTHER: We have 5 cases with pre-treatment and post-treatment biopsy and matched normal tissue and 8 cases with post-treatment biopsy and matched normal tissue. Further 9 cases were collected but the quality of the isolated DNA was not sufficient for next generation sequencing.
A-PREDICT: Patients enrolled in this trial have a pre-treatment and an on-treatment biopsy as well as multiregional sampling of the primary specimen if/when they undergo a nephrectomy. The recruitment into this trial is ongoing. We have received ten cases which consist of a pre-treatment biopsy (sometimes multiple biopsies), and multiple (>10) regions sampled from the nephrectomy specimen as well as the matched normal tissue.
The samples so far have been profiled by whole exome sequencing using the Agilent capture kit and sequencing on the Illumina HiSeq (WORK PACKAGE 2). 56 cases have been profiled to date, of which 10 have been published and data deposited in the European Genome –Phenome archive (EGA) under accession: EGAS00001000667. Gene expression microarray data from the same cases have been deposited in the Gene Expression Omnibus (GEO) archive, under the accession number: GSE53000. For these ten cases multi-regional gene expression studies were also performed revealing that signatures of both good and bad prognosis co-exist in the same tumour further illustrating the limitations of the single-biopsy approach. Whole exome sequencing for the remaining samples as detailed below has either been completed and the data are being analysed or the sequencing is ongoing.
Table 1: Summary of the sequencing of PREDICT samples
Therefore we will eventually have the data on 95 cases including genomic profiling clinical and pathological evaluation and clinical outcome (whenever these can be released). Even taking into account single biopsy data this will be a uniquely comprehensive dataset of RCC cases, which researchers will be able to query with regard to their gene or association of interest or any new therapeutic targets that emerge in due course
Where multiple tumour regions are available we will further inform the ITH in RCC. Where single biopsies are available we will be able to detect clonal biomarkers of therapeutic response (which will be represented in these single biopsies). If such biomarkers are indeed genetic events then with a cohort of almost 100 samples we are powered to detect them.
Establishing xenograft models for renal cancer research (WORK PACKAGE 3): One of the main problems in translational oncology is how to bridge the gap between basic science, molecular biology, cell line base research and actual clinical practice. Xenograft models of actual human tumor biopsies are a promising model system that allow researchers to investigate human tumors that retain as many of the essential biology as possible without posing any risk to patients.
The PREDICT Consortium generated a large number of in vivo xenograft models and in vitro cell bulks from patient derived renal cell carcinoma tumor tissue. 140 nephrectomy specimens from 95 patients were xenotransplanted and 26 RCC xenograft models from 16 patients have been established. The sensitivity of those xenografts have been tested for Sunitinib and Everolimus. Details and availability of those xenografts models can be found on the PREDICT website.
In addition to the xenograft models 43 RCC cell bulks have been generated and characterized. Details and availability of those cell lines is also listed on the consortium website.
Considering the limited number of established RCC cell lines and the value of freshly established tumor cell lines, PREDICT has created a valuable resource for the entire research community.
Isogenic cell lines for renal cancer research (WORK PACKAGE 4):
Isogenic cell lines are a powerful tool to study protein function in the same genetic background. During the course of the PREDICT study we have generated six populations of HK2 cells harbouring genetic modifications of PBRM1. All clones or pool populations obtained have to date been validated at a genomic level. The HK2 PBRM1 KOs were obtained using Zinc finger nucleases, and we have been able to validate them genomically and determine which alleles contain deletions, insertions or SNPs at the targeted locus.
The use of the CRISPR/Cas system to knock out or edit specific genes is becoming more and more common. Within the PREDICT we have generated a Hek293 cell line (Cas9 Nickase (D10A)) that has Cas9D10A stably integrated at the Rosa26 locus (HD 175-006). The benefits of a CRISPR Ready Cell Line is a higher targeting efficiency as only the transfection of the gRNA plasmid/RNA molecule is required. This cell line could be an attractive tool for high throughput screenings using sgRNA libraries.
All the cell lines mentioned above are available upon request.
Predicting individual response and resistance to VEGFR/mTOR pathway therapeutic intervention using biomarkers discovered through tumour functional genomics (WORK PACKAGE 6):
Today, small interfering RNA (siRNA) libraries and vector-based short hairpin RNA (shRNA) libraries are widely used to elucidate gene functions in genome-wide screens by specifically knocking down target gene expression through RNA interference. A special focus of the application of RNAi genetic screens is on improving biomarker and target discovery for personalized treatment of cancer, including the prediction of a patient´s response towards a specific chemotherapeutic regimen. Classically, knockdown constructs of such libraries are designed according to a set of empirically validated algorithms, and they are based on genome sequence data rather than on the transcriptome. Therefore, these libraries cannot reflect any specific pathological situation, e.g. mutations, deletions or gene fusions, nor differences between individual patients. Another major draw-back of the classical libraries is that they target only known RNA transcripts and splice variants. However, as many important RNA structures are fairly unknown, such genome-wide screens may fail to identify the most promising targets.
In order to specifically target RNA expressed in individual RCC patients, a proprietary procedure for the enzymatic generation of transcriptome-based shRNA libraries developed at NMI has been applied within the PREDICT project (WORK PACKAGE 4). Following this protocol a set of 11 immortalized renal cell carcinoma (RCC) cell lines (RCC4, 786-0 A498, 769P, A704, Caki1, Caki2, TK10, UMRC2, UMRC3 and UO31) as well as primary cell lines derived from tumor tissue of 12 RCC patients were stably transduced with patient-derived shRNA (pshRNA) libraries within the course of the project. These pshRNA libraries were produced directly from tumor tissue of the very same patients of which the cell lines have been generated. The procedure of pshRNA library generation encompassed the isolation of RNA from tumor tissue followed by the generation of double-stranded (ds) cDNA and its enzymatic restriction digest. Fragments of 21nt were linked with adaptor and hairpin sequences, amplified and normalized to finally generate a library of pshRNAs reflecting the transcriptome of the individual RCC cell line and patient, respectively. Each individual pshRNA library was then lentivirally bulk transduced into the very RCC tumor cells from which the library was generated. Cryostabilates of bulk pshRNA library transduced RCC cells were shipped to PREDICT partners for comprehensive gene silencing screens and identification of genes involved in everolimus sensitivity and hypoxia sensitivity using clonal sequencing for identification of augmented or depleted knockdown constructs after treatment with everolimus or hypoxia. All pshRNA library transduced cell lines are available to the scientific community via NMI (point of contact: christian.schmees@nmi.de) for screening purposes upon signing of a MTA restricting the use to academic research purposes.
The panel of pshRNA library expressing RCC cell lines generated within PREDICT represents a highly valuable tool for the scientific community to further support the personalized identification of target genes associated with the resistance of individual RCC patients and patient cohorts towards standard-of-care as well as novel therapeutic approaches for this malignancy.
Identification of kinases with altered activities upon tyrosine kinase inhibitor treatment in the discovery cohorts (WORK PACKAGE 5):
We have evaluated the phosphorylation levels of a panel of receptor tyrosine kinase (RTK) in RCC clinical samples after treatment with a multi-tyrosine kinase inhibitor (pazopanib). The phosphorylation levels of the RTKs were assessed using Human Receptor Tyrosine Kinases phosphorylation Antibody Array (RayBioTech). This array allows toidentify simultaneously the relative level of phosphorylation of 71 different human RTK (see list attached).
Our results, as shown in the figure above identified 6 phosphorylated RTK (pRTK) with a higher expression levels in tumor/normal and 6 with a lower expression levels in tumor/normal (p<0.07). These data show that levels of phospho-ZAP70 and phospho-RET were increased in tumor tissue after pazopanib treatment.
ZAP70 is involved in T cell development and lymphocyte activation. We found also high expression levels of pTXK (TCR signal transduction) and pBtk (BCR signalling), two others RTK implicated in immune response signalling. Overall, these findings suggested that pazopanib could mediate activation of immune system. Like the majority of receptor tyrosine kinases studied, signaling pathways initiated by the RET receptor include the Ras/MAP kinase, PI3 kinase/Akt, and phospholipase C-γ (PLCγ) pathways. These pathways are mediators of cellular proliferation and survival.
Identifying genes that modulate sensitivity to Everolimus or hypoxia (WORK PACKAGES 6 and 7)
All therapeutic intervention in ccRCC suffers from a lack of predictive biomarkers, preventing an efficient stratification of patients into responders and non-responders. In order to identify predictive biomarkers we performed whole genome siRNA screens in the absence or presence of Everolimus, an mTOR inhibitor. Our combined screening efforts have identified and validated 13 candidate genes where low expression correlated with reduced drug sensitivity. An additional 33 genes, of which 17 are defined as druggable, were found to sensitize ccRCC cells to Everolimus treatment and could possibly act as new drug targets. The relevance of our findings was substantiated by combining the hits from the screens with genomic analysis of ccRCC tumour samples. This strategy identified multiple genes that are over expressed or amplified in ccRCC, indicating a role in disease progression, a hypothesis that was further strengthened by the finding that the expression of some of these genes also correlated with clinical outcome. Our findings have so far generated one publication(Gerlinger et al., 2013) as well as a manuscript in preparation (Santos et al. 2015) Further work is being done on the additional hits identified in the screens discussed above.
During the course of the PREDICT study, several studies suggested that a mechanism of vascular normalization, as well as tumour secretion of pro-angiogenic factors such as PDGF and FGF, coincide with resistance to anti-angiogenic drugs. These publications suggested that resistance to anti-angiogenic drugs is not mediated by a single resistance mechanism driven by tumour hypoxia, hence our focus within PREDICT shifted to studying the mechanisms of ITH.
As has been indicated above, one of the major scientific breakthroughs generated by the PREDICT consortium is identification of the extensive intratumour heterogeneity (ITH) present in RCC (Gerlinger et al., 2012; 2014)but also in non-small cell lung cancer(de Bruin et al., 2014). Since the initial publication describing ITH as a confounder of precision medicine in ccRCC, we and others have identified ITH in multiple cancers types. It is now widely accepted that a single biopsy will fail to correctly map the mutational landscape of a tumour and that sub populations within a tumour, which could be missed by a single biopsy, may harbor multiple possible resistance mechanisms. This finding has had a profound impact, on not only clinical practice, but has led to a paradigm shift in the field of translational medicine, opened up a new field of study and changed the approach to personalized medicine.
Since the publication of Gerlinger et al. (Gerlinger et al., 2012), ITH has been recognized to be a key to tumour diversity, adaptation and thereby the emergence of drug resistance. This makes the study of tumour evolution, mechanisms of generating ITH as well as identifying common truncal mutations critical for our understanding and treatment of the disease.
ITH can develop through multiple mechanisms such as microsatellite instability (MSI) as well as chromosomal instability and aneuploidy. Within the PREDICT framework, we have been able to identify replication stress as a mechanism for ITH, leading to both structural chromosome aberrations as well as chromosome miss segregation (Burrell et al., 2013).
In addition, long term studies in diploid versus tetraploid cell lines revealed that tetraploid cells undergo rapid genomic changes that mimic genetic alterations identified in chromosomally unstable tumours (Dewhurst et al., 2014).
During our studies we have identified private SETD2 (Su(var), Enhancer of zeste, Trithorax-domain containing 2) mutations within the same tumour, the identification of convergent evolution on this gene highlights the importance of genomic instability on ccRCC tumour development ((Gerlinger et al., 2012), see figure below). SETD2 is a histone trimethylase tri-methylating lysine 36 of histone H3 and multiple mutations within the same tumour indicates a strong selection pressure favoring SETD2 inactivation. Our findings indicated a role for SETD2 in maintaining genome integrity through nucleosome stabilization, the coordination of DNA repair as well as the suppression of replication stress. The study was recently published (Kanu et al., 2015) and was the first time a histone modification was shown to influence replication.
SOFTWARE tools developed by the PREDICT consortium to aid renal cancer research (WORK PACKAGE 8):
A central challenge in genomics is to interpret the mutual functional impact of the large number of molecular abnormalities existent in the same cancer. We developed an on-line software tool, PrediGO (http://gyer1-6.sote.hu/predigo/) to display the complex genomic abnormalities in individual cancers in the biological framework of the gene ontology database. The concept behind PrediGO assumes that each cancer harbors a variable number and unique combination of pathway anomalies and the individual molecular events that underlie the pathway level disturbances vary from case-to-case. In PrediGO we integrate gene expression, copy number alterations and mutation status for each gene to define a complex genomic fingerprint of the investigated tumor. The output plot shows what biological processes are significantly enriched in genomic anomalies in an individual cancer.
Three important aspects of the software have to be noted when comparing to other approaches: our method does not assign functional importance to genes; it is left to the investigators to select the genes for analysis using methods of their choice. Second, we do not build networks from the data but use an external, fixed reference network that reflects functional relationships between genes. Third, we consider all selected genes potentially functionally important and map these into annotated biological pathways to identify biological processes that are heavily affected by molecular anomalies.
PrediGO can be applied to a single case as well: it can be used after upload of the processed data in a pre-defined table format. However, to enable system testing, as an example we have included in the system two of the RCC samples processed by PREDICT. By selecting these samples, the user can immediately grasp the functional set-up of the analysis platform. Color codes mark the different GO terms – grey nodes are not significantly affected, while red nodes have the highest significance. The following example displays a section of the GO network for RCC sample #1 in PredGO with a set of key GO nodes:
OCELOT
Genome-wide molecular measurements of tumor specimens have the potential to provide a wealth of information about the tumor. These measurements can be combined with histological measurements and patient outcome to better understand the molecular mechanisms affecting tumor development, progression, and response or resistance to therapy. However, the analysis involved in such comparisons can be time-consuming and require specialized expertise. We developed a system to manage multiple cancer gene expression data sets, and a web-based tool to quickly analyze this data to perform routine comparisons between gene expression levels and clinical measurements.
The OCELOT (Online Cancer Expression analysis Tool) is an open-source system with two modules: First, the data management module, which is accessible only to the system managers, provides tools to import gene expression data from GEO or from other sources, and to map the associated clinical data to specific clinical terms. Second, the analytical module, which can be open to the public, provides the user with statistically sound tools to assess the association between a particular gene and outcome using Kaplan-Meier survival analysis or ROC curves, or to compare gene expression among various groups, or to compare one gene to another gene. OCELOT performs these tasks in a simple and secure manner, and it can be readily deployed on a basic web server.
Although OCELOT is general enough to work with data from any source, we have focused in particular on renal cancer to develop OCELOT-Renal, a realization of OCELOT with data sets chosen for their relevance to PREDICT research and to renal cancer research in general. OCELOT-Renal can be accessed here: www.cbs.dtu.dk/services/ocelot-1.0/renal/
SEQUENZA
Exome sequencing of tumor specimens is an important component of the PREDICT Consortium’s quest to identify determinants of drug response in renal cancer. In PREDICT research and cancer research in general, exome sequencing is typically used to identify small scale somatic mutations, whereas large scale mutations such as copy number aberrations are typically detected with SNP microarrays. However, exome sequencing has potential to deliver copy number profiles as well, rendering SNP array analysis unnecessary.
Sequenza is a state-of-the-art software to determine tumor copy number profiles directly from matched tumor/normal exome sequencing data. Sequenza provides higher accuracy than other published algorithms in estimation of absolute copy number profile. Furthermore, to the best of our knowledge, it is the only software to produce allele-specific copy number profiles from exome sequencing in a manner that accounts for and models tumor ploidy and cellularity.
Importantly, these accurate copy number profiles enable model-based interpretation of point mutations (variant allele frequency), which in turn enables more accurate detection of mutations, as well as detection of selection among subclonal populations.
Sequenza was developed by members of the PREDICT Consortium and has been freely available to the cancer research community since December 2013. A manuscript describing Sequenza was published in Annals of Oncology (online October 2014, print January 2015).
TUMORTRACER
Recent cancer genome sequencing projects have generated a wealth of insights into the specific genetic defects that lead to tumorigenesis in various types of cancer, as well as the underlying mutational processes that lead to these defects. Several high-profile papers have described pan-cancer comparative analyses of the mutations observed in various cancer types. However, these analyses have generally been descriptive rather than practical.
With our algorithm/software "TumorTracer" we aimed to leverage these important findings to develop a useful diagnostic application. We developed, optimized, and validated a novel, sophisticated algorithm that uses the set of mutations (including copy number changes, when available) detected through genomic analysis of an individual tumor to classify its likely tissue of origin. One important, non-obvious result of our analysis is that classification accuracy is maximized by using a combination of driver mutations, frequencies of various classes of base substitutions that indicate mutational processes, and copy number aberrations.
This algorithm has potential immediate utility in the diagnosis of patients with metastatic cancer of unknown primary (CUP). Similarly, we propose that our algorithm could be used to assist in prioritization of site-specific examinations in patients newly diagnosed with metastatic disease for which the primary site has not yet been located. Furthermore, our method has potential future utility in the early detection of tumors, which may be the most efficient way to reduce the mortality from cancer. As sequencing prices continue to decrease, cancer screening by sequencing of circulating DNA and/or tumor cells seems likely to become widespread, and our algorithm will enable immediate cancer diagnosis at the time of detection.
TumorTracer was developed by members of the PREDICT Consortium. An experimental server implementing the TumorTracer algorithm is available here: http://www.cbs.dtu.dk/services/TumorTracer/.
A manuscript describing TumorTracer has been submitted for publication.
Potential Impact:
The potential impact of the PREDICT consortium can be summarized in three main points:
1) The PREDICT Consortium, through the most cited paper in 2012, has changed the course of cancer research by demonstrating the importance of intratumor heterogeneity and also offers strategies on how to develop effective therapeutic strategies despite ITH
2) The PREDICT Consortium has established valuable resources for various aspects of renal cancer research. Molecular profiles of surgical biopsies, newly established cell lines, xenografts etc. Details and availability of all these resources can be found on the PREDICT consortium website.
3) The PREDICT Consortium has identified several promising biomarkers of treatment response that are currently being validated.
The potential impact (including the socio-economic impact and the wider societal implications of the project so far and the main dissemination activities and exploitation of results
The study results have afforded an unprecedented view of the genetic complexity of renal cell carcinoma demonstrating both inter-individual and intra-individual variation. The socio-economic impact of these findings arises from the appreciation of the need for a personalised approach to each patient and further for that approach to remain fluid rather than fixed along the cancer evolution/patient journey. While in the immediate setting this might have an adverse impact on the cost of cancer management (e.g. profiling of tumours) and treatment (the need to develop new therapies) in the longer run it will avoid both unnecessary treatment costs and toxicities associated with futile therapies. This will result in treatment cost savings but also a reduction in the economic burden of toxicity management. Further, providing precision medicine approach for each patient will result in their improved quality of life and longevity reducing the burden on carers and the society as a whole and possibly enabling some patients to continue contributing to the economy. Where relevant predictive markers are identified a certain number of cases of metastatic diseases are expected to be prevented, thus increasing the number of patients who are able to return to normal life (both family and work) with no adverse effect on life expectancy.
List of Websites:
The website of the consortium can be found at http://www.predictconsortium.eu/.
The consortium generated clinical data, molecular profiles of tumor biopsies, software tools and a wide array of experimental tools for renal cancer research. Those are all available to the community and information about how to obtain those can be found on the consortium website.
The main goal of the PREDICT consortium was to implement a research strategy that could identify predictive biomarkers of response to novel therapeutic interventions in renal cancer. Ours was a comprehensive bed-to-bench-to-bed research strategy that started with clinical samples, analyzed using the most powerful molecular profiling technologies including next generation sequencing. Additional biomarker leads were generated by siRNA based knock-down technologies, which were validated in newly established kidney cancer cell lines as well as isogenic cell lines. Finally, all these data were queried in the validation cohort of clinical samples.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Our first set of results required a profound rethinking of our approach and the ensuing publication has fundamentally changed the course of translational cancer research. As it became evident, kidney cancer shows a significant level of intratumour heterogeneity (ITH). Critically, the majority of known cancer driver events (and therefore potential biomarkers) were subclonal suggesting that they are under-represented in single biopsy approaches. Furthermore, the biological significance of ITH is confirmed by minor subclones that become lethal drug-resistant and metastatic clones. Our SOPs have been altered to capture this heterogeneity using multi-regional tumour sampling. Although we have been able to perform such sampling in a subset of cases for many cases recruited into the studies we only have single biopsy specimens. We have nevertheless embarked on profiling these samples but we have had to adjust our analytical approaches.
Project Context and Objectives:
Effective translational research requires the integration of 1) understanding and clearly formulating the clinical problem at hand; 2) procurement and molecular characterization of appropriate clinical samples; 3) identification and application of the most relevant experimental models systems to the clinical problem; 4) translating the basic research results back into actual clinical practice.
The Objectives of the PREDICT Consortium followed the above described research strategy:
Objective 1: Understanding and clearly formulating the clinical problem at hand. Considering the large average size of primary renal tumours and our awareness of the existing scientific indications of intratumor heterogeneity we had to explore the issue of intratumor heterogeneity in a scientifically rigorous, quantifiable manner.
Objective 2: procurement and molecular characterization of appropriate clinical samples
Objective 2A: procurement of surgical material of primary renal tumours prior to everolimus or TKI treatment
Objective 2B: Sequencing of tumour samples
Objective 2C: Considering the importance of tyrosine kinase activity in the mechanism of effective TKI or everolimus treatment and the known discrepancy between the genomic and kinase activity changes, we designed experiments to directly measure kinase activity in the surgical specimens.
Objective 3: identification and application of the most relevant experimental models systems to the clinical problem
Objective 3A: Considering the limited number of cell lines for renal cancer cell lines available research, we embarked upon establishing novel RCC cell lines.
Objective 3B: Considering the limitations of cell line based experiments to the actual human clinical samples we aimed to establish a valuable panel of xenograft model systems for the renal cancer research community.
Objective 3C: Considering the limitations of associative analysis in biomarker research we established powerful causative discovery strategies by implementing several siRNA screen based research strategies.
Project Results:
Patient samples (WORK PACKAGE 1): One of the most important ingredients of translational research is the establishment of a carefully curated, well documented biobank of appropriately selected patient material. This was achieved to the highest possible standard permitted by changing clinical practice. Our original plans were significantly influenced by the fact that during the course of the PREDICT project the clinical use of mTOR inhibitors, everolimus has significantly changed. Since more effective first line treatments are available for RCC, everolimus had become a 2nd or 3rd line treatment option. This had made it both impossible and unethical to recruit the number of patients originally planned to be treated with everolimus. Yet, prior to this shift in the therapeutic use everolimus, in the context of E-PREDICT and NEORAD clinical trials we recruited 32 patients with careful clinical annotation and comprehensive outcome and follow-up data. We collected multiregionally sampled nephrectomy specimens from all 7 E-PREDICT cases and in two of those cases we also collected matched metastatic tissue. These tumor samples served and will serve at least two important purposes. First, these cases formed the basis of two mile-stone publications that effectively established the field of genomic analysis of intratumor heterogeneity (Gerlinger et al., 2014; 2012). These works have been covered and discussed by several editorials, reviews and we will mention here only the most important findings. While it has been known for a long time that tumor cells show significant heterogeneity the resolution of ITH at single nucleotide resolution was lacking. We applied multiregion whole exome sequencing to resolve the genetic architecture and evolutionary histories of ten ccRCCs. Ultra-deep sequencing identified ITH in all cases. We found that 73-75% of identified ccRCC driver aberrations were subclonal, confounding estimates of driver mutation prevalence. ITH increased with the number of biopsies analyzed, without evidence of saturation in most tumours. Chromosome 3p loss and VHL aberrations were the only ubiquitous events. Multi-regional tumour sampling was also performed on the cases enrolled in the NEORAD trial. We obtained sets for 23 cases consisting of: pre-treatment biopsy, 3 regions from the primary tumour and matched normal. The analysis of these samples is ongoing (see below for time estimates).
PREINSUT: We have 14 cases consisting of pre-treatment and post-treatment biopsy and matched normal tissue. 7 cases could not be included either due to poor quality of the isolated DNA or lack of matched normal tissue. Without the latter genomic data are extremely challenging to interpret, and therefore we proceeded with profiling of 7 cases in total.
SuMR: Of the 14 cases for which tissue was available (consisting of pre-treatment and post-treatment biopsy and matched normal tissue) 13 cases were suitable to sequencing.
SURTIME: We have received 7 cases with a single biopsy of the primary and matched normal. We are expecting to receive material for further 6 cases by June 2015.
PANTHER: We have 5 cases with pre-treatment and post-treatment biopsy and matched normal tissue and 8 cases with post-treatment biopsy and matched normal tissue. Further 9 cases were collected but the quality of the isolated DNA was not sufficient for next generation sequencing.
A-PREDICT: Patients enrolled in this trial have a pre-treatment and an on-treatment biopsy as well as multiregional sampling of the primary specimen if/when they undergo a nephrectomy. The recruitment into this trial is ongoing. We have received ten cases which consist of a pre-treatment biopsy (sometimes multiple biopsies), and multiple (>10) regions sampled from the nephrectomy specimen as well as the matched normal tissue.
The samples so far have been profiled by whole exome sequencing using the Agilent capture kit and sequencing on the Illumina HiSeq (WORK PACKAGE 2). 56 cases have been profiled to date, of which 10 have been published and data deposited in the European Genome –Phenome archive (EGA) under accession: EGAS00001000667. Gene expression microarray data from the same cases have been deposited in the Gene Expression Omnibus (GEO) archive, under the accession number: GSE53000. For these ten cases multi-regional gene expression studies were also performed revealing that signatures of both good and bad prognosis co-exist in the same tumour further illustrating the limitations of the single-biopsy approach. Whole exome sequencing for the remaining samples as detailed below has either been completed and the data are being analysed or the sequencing is ongoing.
Table 1: Summary of the sequencing of PREDICT samples
Therefore we will eventually have the data on 95 cases including genomic profiling clinical and pathological evaluation and clinical outcome (whenever these can be released). Even taking into account single biopsy data this will be a uniquely comprehensive dataset of RCC cases, which researchers will be able to query with regard to their gene or association of interest or any new therapeutic targets that emerge in due course
Where multiple tumour regions are available we will further inform the ITH in RCC. Where single biopsies are available we will be able to detect clonal biomarkers of therapeutic response (which will be represented in these single biopsies). If such biomarkers are indeed genetic events then with a cohort of almost 100 samples we are powered to detect them.
Establishing xenograft models for renal cancer research (WORK PACKAGE 3): One of the main problems in translational oncology is how to bridge the gap between basic science, molecular biology, cell line base research and actual clinical practice. Xenograft models of actual human tumor biopsies are a promising model system that allow researchers to investigate human tumors that retain as many of the essential biology as possible without posing any risk to patients.
The PREDICT Consortium generated a large number of in vivo xenograft models and in vitro cell bulks from patient derived renal cell carcinoma tumor tissue. 140 nephrectomy specimens from 95 patients were xenotransplanted and 26 RCC xenograft models from 16 patients have been established. The sensitivity of those xenografts have been tested for Sunitinib and Everolimus. Details and availability of those xenografts models can be found on the PREDICT website.
In addition to the xenograft models 43 RCC cell bulks have been generated and characterized. Details and availability of those cell lines is also listed on the consortium website.
Considering the limited number of established RCC cell lines and the value of freshly established tumor cell lines, PREDICT has created a valuable resource for the entire research community.
Isogenic cell lines for renal cancer research (WORK PACKAGE 4):
Isogenic cell lines are a powerful tool to study protein function in the same genetic background. During the course of the PREDICT study we have generated six populations of HK2 cells harbouring genetic modifications of PBRM1. All clones or pool populations obtained have to date been validated at a genomic level. The HK2 PBRM1 KOs were obtained using Zinc finger nucleases, and we have been able to validate them genomically and determine which alleles contain deletions, insertions or SNPs at the targeted locus.
The use of the CRISPR/Cas system to knock out or edit specific genes is becoming more and more common. Within the PREDICT we have generated a Hek293 cell line (Cas9 Nickase (D10A)) that has Cas9D10A stably integrated at the Rosa26 locus (HD 175-006). The benefits of a CRISPR Ready Cell Line is a higher targeting efficiency as only the transfection of the gRNA plasmid/RNA molecule is required. This cell line could be an attractive tool for high throughput screenings using sgRNA libraries.
All the cell lines mentioned above are available upon request.
Predicting individual response and resistance to VEGFR/mTOR pathway therapeutic intervention using biomarkers discovered through tumour functional genomics (WORK PACKAGE 6):
Today, small interfering RNA (siRNA) libraries and vector-based short hairpin RNA (shRNA) libraries are widely used to elucidate gene functions in genome-wide screens by specifically knocking down target gene expression through RNA interference. A special focus of the application of RNAi genetic screens is on improving biomarker and target discovery for personalized treatment of cancer, including the prediction of a patient´s response towards a specific chemotherapeutic regimen. Classically, knockdown constructs of such libraries are designed according to a set of empirically validated algorithms, and they are based on genome sequence data rather than on the transcriptome. Therefore, these libraries cannot reflect any specific pathological situation, e.g. mutations, deletions or gene fusions, nor differences between individual patients. Another major draw-back of the classical libraries is that they target only known RNA transcripts and splice variants. However, as many important RNA structures are fairly unknown, such genome-wide screens may fail to identify the most promising targets.
In order to specifically target RNA expressed in individual RCC patients, a proprietary procedure for the enzymatic generation of transcriptome-based shRNA libraries developed at NMI has been applied within the PREDICT project (WORK PACKAGE 4). Following this protocol a set of 11 immortalized renal cell carcinoma (RCC) cell lines (RCC4, 786-0 A498, 769P, A704, Caki1, Caki2, TK10, UMRC2, UMRC3 and UO31) as well as primary cell lines derived from tumor tissue of 12 RCC patients were stably transduced with patient-derived shRNA (pshRNA) libraries within the course of the project. These pshRNA libraries were produced directly from tumor tissue of the very same patients of which the cell lines have been generated. The procedure of pshRNA library generation encompassed the isolation of RNA from tumor tissue followed by the generation of double-stranded (ds) cDNA and its enzymatic restriction digest. Fragments of 21nt were linked with adaptor and hairpin sequences, amplified and normalized to finally generate a library of pshRNAs reflecting the transcriptome of the individual RCC cell line and patient, respectively. Each individual pshRNA library was then lentivirally bulk transduced into the very RCC tumor cells from which the library was generated. Cryostabilates of bulk pshRNA library transduced RCC cells were shipped to PREDICT partners for comprehensive gene silencing screens and identification of genes involved in everolimus sensitivity and hypoxia sensitivity using clonal sequencing for identification of augmented or depleted knockdown constructs after treatment with everolimus or hypoxia. All pshRNA library transduced cell lines are available to the scientific community via NMI (point of contact: christian.schmees@nmi.de) for screening purposes upon signing of a MTA restricting the use to academic research purposes.
The panel of pshRNA library expressing RCC cell lines generated within PREDICT represents a highly valuable tool for the scientific community to further support the personalized identification of target genes associated with the resistance of individual RCC patients and patient cohorts towards standard-of-care as well as novel therapeutic approaches for this malignancy.
Identification of kinases with altered activities upon tyrosine kinase inhibitor treatment in the discovery cohorts (WORK PACKAGE 5):
We have evaluated the phosphorylation levels of a panel of receptor tyrosine kinase (RTK) in RCC clinical samples after treatment with a multi-tyrosine kinase inhibitor (pazopanib). The phosphorylation levels of the RTKs were assessed using Human Receptor Tyrosine Kinases phosphorylation Antibody Array (RayBioTech). This array allows toidentify simultaneously the relative level of phosphorylation of 71 different human RTK (see list attached).
Our results, as shown in the figure above identified 6 phosphorylated RTK (pRTK) with a higher expression levels in tumor/normal and 6 with a lower expression levels in tumor/normal (p<0.07). These data show that levels of phospho-ZAP70 and phospho-RET were increased in tumor tissue after pazopanib treatment.
ZAP70 is involved in T cell development and lymphocyte activation. We found also high expression levels of pTXK (TCR signal transduction) and pBtk (BCR signalling), two others RTK implicated in immune response signalling. Overall, these findings suggested that pazopanib could mediate activation of immune system. Like the majority of receptor tyrosine kinases studied, signaling pathways initiated by the RET receptor include the Ras/MAP kinase, PI3 kinase/Akt, and phospholipase C-γ (PLCγ) pathways. These pathways are mediators of cellular proliferation and survival.
Identifying genes that modulate sensitivity to Everolimus or hypoxia (WORK PACKAGES 6 and 7)
All therapeutic intervention in ccRCC suffers from a lack of predictive biomarkers, preventing an efficient stratification of patients into responders and non-responders. In order to identify predictive biomarkers we performed whole genome siRNA screens in the absence or presence of Everolimus, an mTOR inhibitor. Our combined screening efforts have identified and validated 13 candidate genes where low expression correlated with reduced drug sensitivity. An additional 33 genes, of which 17 are defined as druggable, were found to sensitize ccRCC cells to Everolimus treatment and could possibly act as new drug targets. The relevance of our findings was substantiated by combining the hits from the screens with genomic analysis of ccRCC tumour samples. This strategy identified multiple genes that are over expressed or amplified in ccRCC, indicating a role in disease progression, a hypothesis that was further strengthened by the finding that the expression of some of these genes also correlated with clinical outcome. Our findings have so far generated one publication(Gerlinger et al., 2013) as well as a manuscript in preparation (Santos et al. 2015) Further work is being done on the additional hits identified in the screens discussed above.
During the course of the PREDICT study, several studies suggested that a mechanism of vascular normalization, as well as tumour secretion of pro-angiogenic factors such as PDGF and FGF, coincide with resistance to anti-angiogenic drugs. These publications suggested that resistance to anti-angiogenic drugs is not mediated by a single resistance mechanism driven by tumour hypoxia, hence our focus within PREDICT shifted to studying the mechanisms of ITH.
As has been indicated above, one of the major scientific breakthroughs generated by the PREDICT consortium is identification of the extensive intratumour heterogeneity (ITH) present in RCC (Gerlinger et al., 2012; 2014)but also in non-small cell lung cancer(de Bruin et al., 2014). Since the initial publication describing ITH as a confounder of precision medicine in ccRCC, we and others have identified ITH in multiple cancers types. It is now widely accepted that a single biopsy will fail to correctly map the mutational landscape of a tumour and that sub populations within a tumour, which could be missed by a single biopsy, may harbor multiple possible resistance mechanisms. This finding has had a profound impact, on not only clinical practice, but has led to a paradigm shift in the field of translational medicine, opened up a new field of study and changed the approach to personalized medicine.
Since the publication of Gerlinger et al. (Gerlinger et al., 2012), ITH has been recognized to be a key to tumour diversity, adaptation and thereby the emergence of drug resistance. This makes the study of tumour evolution, mechanisms of generating ITH as well as identifying common truncal mutations critical for our understanding and treatment of the disease.
ITH can develop through multiple mechanisms such as microsatellite instability (MSI) as well as chromosomal instability and aneuploidy. Within the PREDICT framework, we have been able to identify replication stress as a mechanism for ITH, leading to both structural chromosome aberrations as well as chromosome miss segregation (Burrell et al., 2013).
In addition, long term studies in diploid versus tetraploid cell lines revealed that tetraploid cells undergo rapid genomic changes that mimic genetic alterations identified in chromosomally unstable tumours (Dewhurst et al., 2014).
During our studies we have identified private SETD2 (Su(var), Enhancer of zeste, Trithorax-domain containing 2) mutations within the same tumour, the identification of convergent evolution on this gene highlights the importance of genomic instability on ccRCC tumour development ((Gerlinger et al., 2012), see figure below). SETD2 is a histone trimethylase tri-methylating lysine 36 of histone H3 and multiple mutations within the same tumour indicates a strong selection pressure favoring SETD2 inactivation. Our findings indicated a role for SETD2 in maintaining genome integrity through nucleosome stabilization, the coordination of DNA repair as well as the suppression of replication stress. The study was recently published (Kanu et al., 2015) and was the first time a histone modification was shown to influence replication.
SOFTWARE tools developed by the PREDICT consortium to aid renal cancer research (WORK PACKAGE 8):
A central challenge in genomics is to interpret the mutual functional impact of the large number of molecular abnormalities existent in the same cancer. We developed an on-line software tool, PrediGO (http://gyer1-6.sote.hu/predigo/) to display the complex genomic abnormalities in individual cancers in the biological framework of the gene ontology database. The concept behind PrediGO assumes that each cancer harbors a variable number and unique combination of pathway anomalies and the individual molecular events that underlie the pathway level disturbances vary from case-to-case. In PrediGO we integrate gene expression, copy number alterations and mutation status for each gene to define a complex genomic fingerprint of the investigated tumor. The output plot shows what biological processes are significantly enriched in genomic anomalies in an individual cancer.
Three important aspects of the software have to be noted when comparing to other approaches: our method does not assign functional importance to genes; it is left to the investigators to select the genes for analysis using methods of their choice. Second, we do not build networks from the data but use an external, fixed reference network that reflects functional relationships between genes. Third, we consider all selected genes potentially functionally important and map these into annotated biological pathways to identify biological processes that are heavily affected by molecular anomalies.
PrediGO can be applied to a single case as well: it can be used after upload of the processed data in a pre-defined table format. However, to enable system testing, as an example we have included in the system two of the RCC samples processed by PREDICT. By selecting these samples, the user can immediately grasp the functional set-up of the analysis platform. Color codes mark the different GO terms – grey nodes are not significantly affected, while red nodes have the highest significance. The following example displays a section of the GO network for RCC sample #1 in PredGO with a set of key GO nodes:
OCELOT
Genome-wide molecular measurements of tumor specimens have the potential to provide a wealth of information about the tumor. These measurements can be combined with histological measurements and patient outcome to better understand the molecular mechanisms affecting tumor development, progression, and response or resistance to therapy. However, the analysis involved in such comparisons can be time-consuming and require specialized expertise. We developed a system to manage multiple cancer gene expression data sets, and a web-based tool to quickly analyze this data to perform routine comparisons between gene expression levels and clinical measurements.
The OCELOT (Online Cancer Expression analysis Tool) is an open-source system with two modules: First, the data management module, which is accessible only to the system managers, provides tools to import gene expression data from GEO or from other sources, and to map the associated clinical data to specific clinical terms. Second, the analytical module, which can be open to the public, provides the user with statistically sound tools to assess the association between a particular gene and outcome using Kaplan-Meier survival analysis or ROC curves, or to compare gene expression among various groups, or to compare one gene to another gene. OCELOT performs these tasks in a simple and secure manner, and it can be readily deployed on a basic web server.
Although OCELOT is general enough to work with data from any source, we have focused in particular on renal cancer to develop OCELOT-Renal, a realization of OCELOT with data sets chosen for their relevance to PREDICT research and to renal cancer research in general. OCELOT-Renal can be accessed here: www.cbs.dtu.dk/services/ocelot-1.0/renal/
SEQUENZA
Exome sequencing of tumor specimens is an important component of the PREDICT Consortium’s quest to identify determinants of drug response in renal cancer. In PREDICT research and cancer research in general, exome sequencing is typically used to identify small scale somatic mutations, whereas large scale mutations such as copy number aberrations are typically detected with SNP microarrays. However, exome sequencing has potential to deliver copy number profiles as well, rendering SNP array analysis unnecessary.
Sequenza is a state-of-the-art software to determine tumor copy number profiles directly from matched tumor/normal exome sequencing data. Sequenza provides higher accuracy than other published algorithms in estimation of absolute copy number profile. Furthermore, to the best of our knowledge, it is the only software to produce allele-specific copy number profiles from exome sequencing in a manner that accounts for and models tumor ploidy and cellularity.
Importantly, these accurate copy number profiles enable model-based interpretation of point mutations (variant allele frequency), which in turn enables more accurate detection of mutations, as well as detection of selection among subclonal populations.
Sequenza was developed by members of the PREDICT Consortium and has been freely available to the cancer research community since December 2013. A manuscript describing Sequenza was published in Annals of Oncology (online October 2014, print January 2015).
TUMORTRACER
Recent cancer genome sequencing projects have generated a wealth of insights into the specific genetic defects that lead to tumorigenesis in various types of cancer, as well as the underlying mutational processes that lead to these defects. Several high-profile papers have described pan-cancer comparative analyses of the mutations observed in various cancer types. However, these analyses have generally been descriptive rather than practical.
With our algorithm/software "TumorTracer" we aimed to leverage these important findings to develop a useful diagnostic application. We developed, optimized, and validated a novel, sophisticated algorithm that uses the set of mutations (including copy number changes, when available) detected through genomic analysis of an individual tumor to classify its likely tissue of origin. One important, non-obvious result of our analysis is that classification accuracy is maximized by using a combination of driver mutations, frequencies of various classes of base substitutions that indicate mutational processes, and copy number aberrations.
This algorithm has potential immediate utility in the diagnosis of patients with metastatic cancer of unknown primary (CUP). Similarly, we propose that our algorithm could be used to assist in prioritization of site-specific examinations in patients newly diagnosed with metastatic disease for which the primary site has not yet been located. Furthermore, our method has potential future utility in the early detection of tumors, which may be the most efficient way to reduce the mortality from cancer. As sequencing prices continue to decrease, cancer screening by sequencing of circulating DNA and/or tumor cells seems likely to become widespread, and our algorithm will enable immediate cancer diagnosis at the time of detection.
TumorTracer was developed by members of the PREDICT Consortium. An experimental server implementing the TumorTracer algorithm is available here: http://www.cbs.dtu.dk/services/TumorTracer/.
A manuscript describing TumorTracer has been submitted for publication.
Potential Impact:
The potential impact of the PREDICT consortium can be summarized in three main points:
1) The PREDICT Consortium, through the most cited paper in 2012, has changed the course of cancer research by demonstrating the importance of intratumor heterogeneity and also offers strategies on how to develop effective therapeutic strategies despite ITH
2) The PREDICT Consortium has established valuable resources for various aspects of renal cancer research. Molecular profiles of surgical biopsies, newly established cell lines, xenografts etc. Details and availability of all these resources can be found on the PREDICT consortium website.
3) The PREDICT Consortium has identified several promising biomarkers of treatment response that are currently being validated.
The potential impact (including the socio-economic impact and the wider societal implications of the project so far and the main dissemination activities and exploitation of results
The study results have afforded an unprecedented view of the genetic complexity of renal cell carcinoma demonstrating both inter-individual and intra-individual variation. The socio-economic impact of these findings arises from the appreciation of the need for a personalised approach to each patient and further for that approach to remain fluid rather than fixed along the cancer evolution/patient journey. While in the immediate setting this might have an adverse impact on the cost of cancer management (e.g. profiling of tumours) and treatment (the need to develop new therapies) in the longer run it will avoid both unnecessary treatment costs and toxicities associated with futile therapies. This will result in treatment cost savings but also a reduction in the economic burden of toxicity management. Further, providing precision medicine approach for each patient will result in their improved quality of life and longevity reducing the burden on carers and the society as a whole and possibly enabling some patients to continue contributing to the economy. Where relevant predictive markers are identified a certain number of cases of metastatic diseases are expected to be prevented, thus increasing the number of patients who are able to return to normal life (both family and work) with no adverse effect on life expectancy.
List of Websites:
The website of the consortium can be found at http://www.predictconsortium.eu/.
The consortium generated clinical data, molecular profiles of tumor biopsies, software tools and a wide array of experimental tools for renal cancer research. Those are all available to the community and information about how to obtain those can be found on the consortium website.
