REVIEWS Translational research in oncology—10 years of progress and future prospects James H. Doroshow and Shivaani Kummar Abstract | International efforts to sequence the genomes of various human cancers have been broadly deployed in drug discovery programmes. Diagnostic tests that predict the value of the molecularly targeted anticancer agents used in such programmes are conceived and validated in parallel with new smallmolecule treatments and immunotherapies. This approach has been aided by better preclinical cancer models; an enhanced appreciation of the complex interactions that exist between tumour cells and their microenvironment; the elucidation of interactions between many of the genetic drivers of cancer, including oncogenes and tumour suppressors; and recent insights into the genetic heterogeneity of human tumours made possible by extraordinary improvements in DNA-sequencing techniques. These advances are being employed in the first generation of genomic clinical trials that will examine the feasibility of matching a broad range of systemic therapies to specific molecular tumour characteristics. More-extensive molecular characterization of tumours and their supporting matrices are anticipated to become standard aspects of oncological practice, permitting continuous molecular re-evaluations of human malignancies on a patient-bypatient and treatment-by-treatment basis. We review selected developments in translational cancer biology, diagnostics, and therapeutics that have occurred over the past decade and offer our thoughts on future prospects for the next few years. Doroshow, J. H. & Kummar, S. Nat. Rev. Clin. Oncol. advance online publication 7 October 2014; doi:10.1038/nrclinonc.2014.158
Introduction
Division of Cancer Treatment and Diagnosis (J.H.D., S.K.), Developmental Therapeutics Branch of the Center for Cancer Research (J.H.D.), Room 3A‑44, Building 31, 31 Center Drive, National Cancer Institute, NIH, Bethesda, MD 20892, USA.
The past decade has witnessed a remarkable acceler ation in the pace of translational cancer medicine.1–3 The first issue of Nature Clinical Practice Oncology, the forerunner of Nature Reviews Clinical Oncology, which was published 10 years ago, focused on the dramatic clinical benefit of treatment with the EGFR inhibitor gefitinib for patients with non-small-cell lung cancers (NSCLCs) harbouring mutations in the EGFR gene.4–7 The demonstration that the mutational status of a solid tumour could predict therapeutic efficacy for a specific agent in a molecularly-defined subset of patients galva nized oncology drug-development programmes around a new biological paradigm, solidifying the shift toward the elaboration of molecularly targeted therapeutics, highlighted by the introduction of trastuzumab and imatinib into the clinic. These findings ended the era of nonspecific cytotoxin development.8,9 Advances in translational oncology over the past 10 years have been characterized by the application of ever-more-sophisticated molecular tools to larger popula tions of patients with cancer or those who are at increased risk of developing the disease (Figure 1 (Timeline)). These advances include the demonstration that risk of recur rence for women with oestrogen receptor (ER)-positive breast cancer and histologically uninvolved lymph nodes receiving tamoxifen could be predicted based on
Correspondence to: J.H.D. [email protected]
Competing interests The authors declare no competing interests.
the expression levels of 21 genes, determined using a reverse-transcription-PCR (RT‑PCR) assay performed on formalin-fixed tumour tissue (Oncotype DX®, Genomic Health Inc., Redwood City, CA). This test was one of the first fully-validated predictive cancer biomarkers.10 The application of molecular characterization technologies (such as modern tumour tissue acquisition methods, next-generation DNA sequencing, gene-expression analy sis, DNA methylation profiling, proteomic evaluation, and development of ‘big data’ sets) to the understanding of cancer biology in the clinic, exemplified by the work of The Cancer Genome Atlas (TCGA) project 11 and the International Cancer Genome Consortia,12 as well as the application of large-scale human tumour cell line drug screening,13–15 has dramatically expanded appreciation of the broad range of specific mutations and other molecular abnormalities that could be examined as potential targets for therapeutics development. In this way, molecularly targeted therapies for additional subsets of patients with adenocarcinoma of the lung (with rearrangements in the anaplastic lymphoma kinase [ALK] gene) as well as a major portion of the metastatic melanoma patient popu lation (carrying V600E mutations in BRAF) were identi fied; these therapies were then rapidly translated into clinical trials demonstrating substantial therapeutic activ ity for new drugs targeting specific molecular lesions.16,17 Improvements in DNA sequencing also provided the basis for a much more sophisticated appreciation of the evolution of human tumour heterogeneity.18
NATURE REVIEWS | CLINICAL ONCOLOGY
ADVANCE ONLINE PUBLICATION | 1 © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS Translational cancer biology
Key points
Around 10 years ago, the possibility that all of the genes in the exomes of a human cancer could be sequenced with precision and at modest cost was considered an aspirational goal; the fact that such DNA sequencing procedures are now a routine part of translational onco logy research speaks to the rapidly changing landscape of cancer biology.28–30 In concert with advances in DNA sequencing, rapid advances in our understanding of the biology of cancer have occurred since 2004, including major efforts to improve both the model systems and the scientific frameworks used to study human cancers and cancer therapeutics.
■■ The landscape of translational oncology has shifted dramatically over the past 10 years, characterized by the introduction of ever-more-sophisticated molecular tools into the clinic ■■ Translational cancer-biology studies have markedly improved preclinical models applicable for therapeutics development, as well as our understanding of the roles of inflammation and altered intermediary metabolism in carcinogenesis ■■ Translational cancer diagnostics and therapeutics have been revolutionized by the molecular characterization of human tumours, a process that now underlies the development of molecularly-targeted, rather than broadly cytotoxic, anticancer therapies ■■ Improvements in molecular tumour-classification techniques will permit their widespread application for patients at diagnosis, disease recurrence, and during therapy, supporting continuous adaptation of therapeutic approaches to evolving tumour characteristics
New molecular models of cancer The availability of immunodeficient strains of mice, 30 years ago, made it possible to grow established human tumour cell lines as xenotransplants, facilitat ing the testing of cancer therapeutic agents in vivo against a much wider range of tumour histologies than had been possible previously with the use of syngeneic mouse tumour models.31 However, although useful for the demonstration of the potential therapeutic index of a novel drug, years of experience with these models indi cated that they did not faithfully predict histologically- specific antitumour efficacy, particularly for molecularly targeted drugs. This lack of prediction is perhaps a result of the use of tumour cell lines (from which the xenografts are often established) that had been propa gated under the selective pressure of continuous two- dimensional tissue culture for, in some cases, decades.31,32 For this reason, substantive investigational effort has been focused on the development of new in vivo and in vitro models for cancer discovery research. Using these new models, as well as clinical specimens, a better framework for understanding the roles of inflammation and metabolism in carcinogenesis has also developed over the past decade.
The rapidly expanding knowledge of human DNA repair processes has been exploited to develop inhibi tors of certain DNA repair proteins. For example, drugs targeting poly(ADP‑ribose) polymerase (PARP) have shown clinical benefit in women with ovarian cancers who carry germline mutations in the BRCA1 or BRCA2 genes.19,20 Enhanced understanding of the control of tumour cell immunity, and technological advances in the molecular engineering of immune cells, has recently been translated into new immunotherapeutic treatment programmes, including antibodies and cell vaccines, that are beginning to transform cancer therapy.21–23 Finally, tantalizing improvements over the past 10 years in the development of both in vitro and in vivo models of human cancer,24,25 including those produced directly from a specific individual’s tumour,26,27 suggest that in the future we might be able to examine tumour biology on a patient-by-patient basis. In this Review, we evaluate selected developments in translational cancer biology, diagnostics, and therapeutics that have occurred over the past decade; such advances will be used as the basis for suggesting specific prospects for future progress in translational cancer medicine. Lung cancer EGFR mutations and erlotinib sensitivity
2004
Sensitivity of BRCA mutant tumours to poly (ADP-ribose) polymerase inhibitors
2006
Oncotype Dx™
TCGA begins
Large-scale cell line screening identifies ALK fusion inhibitor for NSCLC
2008
Glioblastoma multiforme genome completed Molecularly targeted melanoma therapy
2010
Immunotherapy: checkpoint inhibitors; genetically engineered T cells
2012
Intratumoral genomic heterogeneity
2014
Development of individually derived tumour models from circulating tumour cells for prospective drug testing
2016
TCGA completed >10,000 tumours
Large-scale screening for ‘actionable’ mutations in adult malignancies, and use of circulating tumour DNA to assess cancer burden and therapeutic efficacy
Figure 1 | Timeline of 10-year translational research for oncology and prospects for the future. Over the past decade, translational cancer research has undergone an extraordinary transformation, spanning the time from the first demonstration of a single, functional driver mutation in a human solid tumour (NSCLC) with clear therapeutic consequences (enhanced response to EGFR inhibitors) to the application of next-generation whole-exome sequencing for therapeutic decision making in a growing population of cancer patients. Future prospects (shown in pink) include the expected completion date for TCGA project and the rapid development of personal human tumour models that may begin to be used for testing of cancer therapeutic agents on an individualized basis. Abbreviations: ALK, anaplastic lymphoma kinase; NSCLC, non-small-cell lung cancer; TCGA, The Cancer Genome Atlas.
2 | ADVANCE ONLINE PUBLICATION
www.nature.com/nrclinonc © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS a GEM
Treat with most appropriate targeted drugs
Tumorigenesis
b GDA Transplantation into syngeneic immunocompetent mice
c PDX and conditionally
reprogrammed cell lines Create reprogrammed cell lines
‘Pre-clinical’ clinical trials Molecularly characterize, treat/screen mice bearing transplants and cells with relevant drugs
Tumorigenesis Transplantation into NSG mice
Tumour/patient heterogeneity
Figure 2 | Improving tumour models for cancer biology and drug development. a | GEMs of human cancer developed through the organ-specific expression of driver mutations in immunocompetent mice, are used to evaluate the signal transduction pathways associated with all stages of carcinogenesis as well as to examine the effects of molecularly targeted agents in proof-of-mechanism studies. b | GDAs facilitate the use of larger numbers of animals carrying tumours that have been biologically defined and derived for pre-clinical clinical trials in mice with intact immune systems. c | PDXs are developed directly from implantation of surgical, biopsy, or circulating tumour cell specimens into immune-incompetent mice; compared with older xenograft models derived from established human tumour cell lines, PDX tumours usually carry with them some degree of human stroma during initial tumour passages in vivo, facilitating a more accurate environment for molecular characterization and drug testing. New in vitro techniques to develop early passage conditionally reprogrammed cell lines or tumour cell organoid cultures are also under active investigation as preclinical models that might more accurately predict the efficacy of targeted therapeutic agents. We acknowledge the important role of Terry van Dyke, National Cancer Institute, NIH, in the development of this figure. Abbreviations: GDA, GEM-derived allografts; GEM, genetically-engineered mouse model; NGS, NOD scid gamma; PDX, patient-derived tumour xenograft model.
Model organisms Among the most important developments in trans lational cancer biology during the past 10 years has been the use of genetic approaches to produce moresophisticated models of human cancer.33,34 For example, the elucidation of critical signalling pathways essen tial for understanding the development of pancreatic ductal adenocarcinoma has been facilitated greatly by genetically-engineered mouse models (GEMs) of this disease that recapitulate the carcinogenic process in mice with an intact immune system.35 The genetic program ming of organ-specific expression of driver mutations not only underlies the development of premalignant lesions and primary tumours in these mice, but allows for proof-of-mechanism studies of targeted therapeu tic agents active against the genetic lesions produced
in such animals (Figure 2a). This exemplifies how the exploration of human tumour biology in model systems has been enhanced through the use of GEMs. Recent efforts to improve the clinical use of GEMs, which have been hindered by the prolonged timelines (often 9–12 months) and expense required to generate the models, have focused on the possibility that orthotopic or subcutaneous transplantation of tumours initially established in GEMs could facilitate their use for drug screening (Figure 2b). GEM allografts may enhance the application of clinical trial design principles in ‘preclini cal clinical trials’ so that these models could more easily be employed with adequate sample sizes, predefined therapeutic hypotheses, and integrated pharmaco dynamic and pharmacokinetic studies, facilitating comparisons of drug efficacy in tumours from GEMs
NATURE REVIEWS | CLINICAL ONCOLOGY
ADVANCE ONLINE PUBLICATION | 3 © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS with human clinical trial results of the same agents and combinations.34 Additional progress will expand the diversity of GEM models to characterize more fully the wide range of molecular aberrations observed in clinical tumour specimens; there is also little question that greater use of GEM technology could assist efforts focusing on biomarker development for early stage disease as well as the evolution of therapeutic resistance. There has also been a dramatic resurgence of interest in mouse tumour models generated directly from patient surgical or biopsy samples, denoted as patient-derived tumour xenografts (PDXs). 36 Although such model systems had been investigated for decades, the availabil ity of NOD/SCID and NOD/SCID/IL2Rγnull (NSG) mice over the past 10 years, which enhances murine immunoincompetence by blocking natural killer cell maturation and facilitates tumour engraftment, has led to a renais sance in the use of this technique for generating tumour models. Furthermore, for certain solid tumours, such as colorectal, pancreatic, lung, and breast cancers, early tumour passages (probably through the third or fourth generation in most cases) retain a degree of human stroma, facilitating studies of gene expression, drug effi cacy, and tumour heterogeneity.36 In the case of endo crine sensitive and resistant breast cancer, PDX models have been used to demonstrate the stability of genomewide allele frequencies, and response to endocrine therapy, between the initial human tumour specimens and PDXs established from these tissues.37 Therefore, the development of large panels of clinically-annotated tumours could help evaluate the diversity of responses expected in the clinical arena, or could be used for molecularly-characterized preclinical trials to assess the validity of target selection strategies in advance of clini cal testing (Figure 2c). These models avoid the problems associated with in vitro tumour cell selection; however, the absence of an intact immune system negates their value for testing immunotherapeutic strategies. Most PDX models have been developed from surgical speci mens rather than from biopsies of metastatic sites; thus, these models might not be representative of the tumours that occur in patients with recurrent malignancies who require systemic therapy.38 Further development of PDX models for both biological studies and drug selection might have an important role in improving the precision with which cancer treatments are selected in the future. While there has been considerable focus on improv ing in vivo models of human cancer, recent studies have suggested better ways to study human tumours in vitro. The immortalization of epithelial cells, including human tumour cells, by inhibiting terminal differentiation through the use of an inhibitor of Rho kinase and con ditioned media, has excited investigational interest; if widely applicable, this ‘conditional reprogramming’ tech nique might facilitate the long-term culture of tumour cells that heretofore have been difficult to propagate, such as tumours of the prostate (Figure 2c).39,40 It is not known whether tumour cells cultured in this fashion undergo genetic adaptation to growth on plastic, which would diminish their predictive potential. The development of
3D cultures of cancer organoids—tumour cells that grow in a collagen-rich matrix supported by essential growth factors and that have phenotypic characteristics of cancer stem cells—is the other major focus for the development of in vitro cancer models. Although these models lack stroma and require very well-defined growth conditions, organoids can be propagated indefinitely from indivi dual patients, are amenable to drug screening, and can retain essential molecular characteristics of the primary disease (at least for colon cancer).41,42 There is great inter est in the possibility that organoid cultures developed for other malignancies (such as pancreatic and prostate cancers) will provide important new in vitro tools for cancer biology and drug discovery.41,42 Inflammation and cancer One of the most important translational research relationships to understand is the interaction between innate inflammatory responses, host defence, and carcinogenesis; our understanding of this relationship, including the role of the human microbiome in these pro cesses, has expanded dramatically over the past decade.43 A deeper appreciation of the molecular mechanisms underlying chronic inflammation-related cancers (such as those associated with inflammatory bowel disease, chronic pancreatitis, and viral hepatitis) and the evidence for the protective role of non-steroidal anti-inflammatory agents in the prevention of multiple malignant histolo gies, indicates that greater emphasis on the development of new therapeutic modalities to prevent the adverse con sequences of the unchecked cytokine release and reactive oxygen production that accompany chronic inflamma tory stress are urgently required. Such therapeutics might interdict the proinflammatory microenvironment that is conducive to DNA damage, cancer initiation, and pro gression.44–47 Recent evidence has demonstrated previ ously unknown relationships between the inflammatory environment, produced by commensal bacteria, and the efficacy of both immune and cytotoxic anticancer thera pies in vivo, emphasizing the role of the tumour micro environment in the control of therapeutic response.48 It seems likely that in the near future a better understand ing of how to control both proinflammatory and antiinflammatory responses will play a critical part in the developmental therapeutics of cancer. Cancer metabolism Another remarkable change in translational cancer biology over the past decade has been the recognition that the metabolic reprogramming of tumour cells is inti mately related to oncogene-induced proliferative signal ling and the function of tumour suppressor genes.49,50 Thus, signal transduction pathways that are important for malignant transformation (for example, those controlled by Myc, p53, and mTOR) have a critical role in altering the metabolic phenotypes of cancer cells, while at the same time the products of altered tumour cell metabolism (α-ketoglutarate and reactive oxygen species [ROS]) affect tumour cell signal transduction.51 To enhance cell pro liferation, tumour cells adapt their metabolic machinery
4 | ADVANCE ONLINE PUBLICATION
www.nature.com/nrclinonc © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS to use a variety of energy sources besides glucose (such as glutamine) to increase the production of fatty acids and other macromolecules needed for the production of cell membranes as well as a diverse portfolio of anabolic processes, in addition to ATP production. The metabolic programmes of tumour cells are diverse, tissue context dependent, highly influenced by the surrounding micro environment as well as the tumour host, and hetero geneous even within a single tumour.52 Recent evidence suggests that metabolic reprogramming of tumour cells might have a critical role in the epigenetic modification of tumour cell DNA and histones.53 For these reasons, substantial investigational effort is now being directed toward the development of small-molecule inhibitors of specific metabolic enzymes that contribute to tumour cell proliferation (such as isocitrate dehydrogenase 1 [IDH1] and IDH2) for low-grade gliomas and acute myelogenous leukaemia, respectively.54,55
Characterization of tumours and host tissues About the same time that imatinib was found to be an effective treatment for chronic myelogenous leukaemia and erlotinib for adenocarcinoma of the lung with EGFR mutations, the sequencing of the human genome was completed.56 The completion of this project propelled the application of genomic sequencing technologies in cancer biology (Figure 1). Although much effort has been focused on the presence of somatic mutations in specific cancer histologies,57 the availability of powerful sequencing techniques also led to a series of populationbased studies to discover germline mutations that could confer cancer risk, so-called genome-wide association studies.58 To date over 150 regions of the human genome, often in noncoding domains, have been correlated with two dozen human cancers. A major area of current investigation is to understand how these genetic vari ants predispose to cancer, and whether or how inter actions occur between somatic and germline variants in cancer development.59 Human biospecimens Molecular tumour characterization, whether it invol ves DNA sequencing, evaluation of RNA expression, proteomic or phosphoproteomic determinations, or immunohistochemistry, requires the collection of high quality biospecimens.60 The lack of accepted national guidelines for specimen acquisition stimulated a multiyear process leading to the introduction of standards for the collection and banking of both tumour and normal tissues.61 Lack of standard biospecimen collec tion, processing, and storage procedures hampered the initiation of the TCGA project; difficulties were encoun tered in obtaining tumour specimens that met quality standards for sequencing in terms of the tumour cell purity of the samples; the fixation method employed; and available volumes of high-quality tumours across several histologies. The value of proof-of-mechanism pharmacodynamic assays is frequently diminished because insufficient care is taken to understand the stability of the analyte of interest
in the context of performing research tumour biopsies; sample handling techniques that will stabilize biomarkers after tissues are acquired are infrequently evaluated before study initiation.62–64 Unfortunately, regardless of the bio marker, there is a need to improve the details of sample acquisition that are fundamental to successful molecular characterization of a tumour or normal tissue of interest. Clinical characterization of tumours The availability of massively parallel DNA sequencing in the mid‑2000s allowed for an exponential increase in the speed and amount of tumour DNA sequence that could be obtained in a given period of time; this enhanced efficiency led to plummeting costs for acquiring DNA sequences and the development of better tools to analyse sequence data.28 Systematic DNA sequencing substan tiated the role of receptor tyrosine kinase mutations in driving tumour cell signalling across many cancers (PI3KCA in breast cancer and JAK2 in myeloprolifera tive syndromes) and has been essential for stimulating the development of new kinase inhibitors for the treat ment of tumours that are dependent on such pathways. The mutational patterns observed—extremely high fre quencies in lung cancer and melanoma, and relatively low levels in untreated paediatric malignancies—have also substantiated aetiological associations with the extent and duration of mutagen exposure in the former malignancies, and have suggested hypotheses for the development of tumours in non-renewing tissue types for the latter.65 A recent analysis indicates that despite the intensive investigation of >10,000 tumours by the TCGA, additional novel driver mutations are likely to be found at low frequency (
Introduction
Division of Cancer Treatment and Diagnosis (J.H.D., S.K.), Developmental Therapeutics Branch of the Center for Cancer Research (J.H.D.), Room 3A‑44, Building 31, 31 Center Drive, National Cancer Institute, NIH, Bethesda, MD 20892, USA.
The past decade has witnessed a remarkable acceler ation in the pace of translational cancer medicine.1–3 The first issue of Nature Clinical Practice Oncology, the forerunner of Nature Reviews Clinical Oncology, which was published 10 years ago, focused on the dramatic clinical benefit of treatment with the EGFR inhibitor gefitinib for patients with non-small-cell lung cancers (NSCLCs) harbouring mutations in the EGFR gene.4–7 The demonstration that the mutational status of a solid tumour could predict therapeutic efficacy for a specific agent in a molecularly-defined subset of patients galva nized oncology drug-development programmes around a new biological paradigm, solidifying the shift toward the elaboration of molecularly targeted therapeutics, highlighted by the introduction of trastuzumab and imatinib into the clinic. These findings ended the era of nonspecific cytotoxin development.8,9 Advances in translational oncology over the past 10 years have been characterized by the application of ever-more-sophisticated molecular tools to larger popula tions of patients with cancer or those who are at increased risk of developing the disease (Figure 1 (Timeline)). These advances include the demonstration that risk of recur rence for women with oestrogen receptor (ER)-positive breast cancer and histologically uninvolved lymph nodes receiving tamoxifen could be predicted based on
Correspondence to: J.H.D. [email protected]
Competing interests The authors declare no competing interests.
the expression levels of 21 genes, determined using a reverse-transcription-PCR (RT‑PCR) assay performed on formalin-fixed tumour tissue (Oncotype DX®, Genomic Health Inc., Redwood City, CA). This test was one of the first fully-validated predictive cancer biomarkers.10 The application of molecular characterization technologies (such as modern tumour tissue acquisition methods, next-generation DNA sequencing, gene-expression analy sis, DNA methylation profiling, proteomic evaluation, and development of ‘big data’ sets) to the understanding of cancer biology in the clinic, exemplified by the work of The Cancer Genome Atlas (TCGA) project 11 and the International Cancer Genome Consortia,12 as well as the application of large-scale human tumour cell line drug screening,13–15 has dramatically expanded appreciation of the broad range of specific mutations and other molecular abnormalities that could be examined as potential targets for therapeutics development. In this way, molecularly targeted therapies for additional subsets of patients with adenocarcinoma of the lung (with rearrangements in the anaplastic lymphoma kinase [ALK] gene) as well as a major portion of the metastatic melanoma patient popu lation (carrying V600E mutations in BRAF) were identi fied; these therapies were then rapidly translated into clinical trials demonstrating substantial therapeutic activ ity for new drugs targeting specific molecular lesions.16,17 Improvements in DNA sequencing also provided the basis for a much more sophisticated appreciation of the evolution of human tumour heterogeneity.18
NATURE REVIEWS | CLINICAL ONCOLOGY
ADVANCE ONLINE PUBLICATION | 1 © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS Translational cancer biology
Key points
Around 10 years ago, the possibility that all of the genes in the exomes of a human cancer could be sequenced with precision and at modest cost was considered an aspirational goal; the fact that such DNA sequencing procedures are now a routine part of translational onco logy research speaks to the rapidly changing landscape of cancer biology.28–30 In concert with advances in DNA sequencing, rapid advances in our understanding of the biology of cancer have occurred since 2004, including major efforts to improve both the model systems and the scientific frameworks used to study human cancers and cancer therapeutics.
■■ The landscape of translational oncology has shifted dramatically over the past 10 years, characterized by the introduction of ever-more-sophisticated molecular tools into the clinic ■■ Translational cancer-biology studies have markedly improved preclinical models applicable for therapeutics development, as well as our understanding of the roles of inflammation and altered intermediary metabolism in carcinogenesis ■■ Translational cancer diagnostics and therapeutics have been revolutionized by the molecular characterization of human tumours, a process that now underlies the development of molecularly-targeted, rather than broadly cytotoxic, anticancer therapies ■■ Improvements in molecular tumour-classification techniques will permit their widespread application for patients at diagnosis, disease recurrence, and during therapy, supporting continuous adaptation of therapeutic approaches to evolving tumour characteristics
New molecular models of cancer The availability of immunodeficient strains of mice, 30 years ago, made it possible to grow established human tumour cell lines as xenotransplants, facilitat ing the testing of cancer therapeutic agents in vivo against a much wider range of tumour histologies than had been possible previously with the use of syngeneic mouse tumour models.31 However, although useful for the demonstration of the potential therapeutic index of a novel drug, years of experience with these models indi cated that they did not faithfully predict histologically- specific antitumour efficacy, particularly for molecularly targeted drugs. This lack of prediction is perhaps a result of the use of tumour cell lines (from which the xenografts are often established) that had been propa gated under the selective pressure of continuous two- dimensional tissue culture for, in some cases, decades.31,32 For this reason, substantive investigational effort has been focused on the development of new in vivo and in vitro models for cancer discovery research. Using these new models, as well as clinical specimens, a better framework for understanding the roles of inflammation and metabolism in carcinogenesis has also developed over the past decade.
The rapidly expanding knowledge of human DNA repair processes has been exploited to develop inhibi tors of certain DNA repair proteins. For example, drugs targeting poly(ADP‑ribose) polymerase (PARP) have shown clinical benefit in women with ovarian cancers who carry germline mutations in the BRCA1 or BRCA2 genes.19,20 Enhanced understanding of the control of tumour cell immunity, and technological advances in the molecular engineering of immune cells, has recently been translated into new immunotherapeutic treatment programmes, including antibodies and cell vaccines, that are beginning to transform cancer therapy.21–23 Finally, tantalizing improvements over the past 10 years in the development of both in vitro and in vivo models of human cancer,24,25 including those produced directly from a specific individual’s tumour,26,27 suggest that in the future we might be able to examine tumour biology on a patient-by-patient basis. In this Review, we evaluate selected developments in translational cancer biology, diagnostics, and therapeutics that have occurred over the past decade; such advances will be used as the basis for suggesting specific prospects for future progress in translational cancer medicine. Lung cancer EGFR mutations and erlotinib sensitivity
2004
Sensitivity of BRCA mutant tumours to poly (ADP-ribose) polymerase inhibitors
2006
Oncotype Dx™
TCGA begins
Large-scale cell line screening identifies ALK fusion inhibitor for NSCLC
2008
Glioblastoma multiforme genome completed Molecularly targeted melanoma therapy
2010
Immunotherapy: checkpoint inhibitors; genetically engineered T cells
2012
Intratumoral genomic heterogeneity
2014
Development of individually derived tumour models from circulating tumour cells for prospective drug testing
2016
TCGA completed >10,000 tumours
Large-scale screening for ‘actionable’ mutations in adult malignancies, and use of circulating tumour DNA to assess cancer burden and therapeutic efficacy
Figure 1 | Timeline of 10-year translational research for oncology and prospects for the future. Over the past decade, translational cancer research has undergone an extraordinary transformation, spanning the time from the first demonstration of a single, functional driver mutation in a human solid tumour (NSCLC) with clear therapeutic consequences (enhanced response to EGFR inhibitors) to the application of next-generation whole-exome sequencing for therapeutic decision making in a growing population of cancer patients. Future prospects (shown in pink) include the expected completion date for TCGA project and the rapid development of personal human tumour models that may begin to be used for testing of cancer therapeutic agents on an individualized basis. Abbreviations: ALK, anaplastic lymphoma kinase; NSCLC, non-small-cell lung cancer; TCGA, The Cancer Genome Atlas.
2 | ADVANCE ONLINE PUBLICATION
www.nature.com/nrclinonc © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS a GEM
Treat with most appropriate targeted drugs
Tumorigenesis
b GDA Transplantation into syngeneic immunocompetent mice
c PDX and conditionally
reprogrammed cell lines Create reprogrammed cell lines
‘Pre-clinical’ clinical trials Molecularly characterize, treat/screen mice bearing transplants and cells with relevant drugs
Tumorigenesis Transplantation into NSG mice
Tumour/patient heterogeneity
Figure 2 | Improving tumour models for cancer biology and drug development. a | GEMs of human cancer developed through the organ-specific expression of driver mutations in immunocompetent mice, are used to evaluate the signal transduction pathways associated with all stages of carcinogenesis as well as to examine the effects of molecularly targeted agents in proof-of-mechanism studies. b | GDAs facilitate the use of larger numbers of animals carrying tumours that have been biologically defined and derived for pre-clinical clinical trials in mice with intact immune systems. c | PDXs are developed directly from implantation of surgical, biopsy, or circulating tumour cell specimens into immune-incompetent mice; compared with older xenograft models derived from established human tumour cell lines, PDX tumours usually carry with them some degree of human stroma during initial tumour passages in vivo, facilitating a more accurate environment for molecular characterization and drug testing. New in vitro techniques to develop early passage conditionally reprogrammed cell lines or tumour cell organoid cultures are also under active investigation as preclinical models that might more accurately predict the efficacy of targeted therapeutic agents. We acknowledge the important role of Terry van Dyke, National Cancer Institute, NIH, in the development of this figure. Abbreviations: GDA, GEM-derived allografts; GEM, genetically-engineered mouse model; NGS, NOD scid gamma; PDX, patient-derived tumour xenograft model.
Model organisms Among the most important developments in trans lational cancer biology during the past 10 years has been the use of genetic approaches to produce moresophisticated models of human cancer.33,34 For example, the elucidation of critical signalling pathways essen tial for understanding the development of pancreatic ductal adenocarcinoma has been facilitated greatly by genetically-engineered mouse models (GEMs) of this disease that recapitulate the carcinogenic process in mice with an intact immune system.35 The genetic program ming of organ-specific expression of driver mutations not only underlies the development of premalignant lesions and primary tumours in these mice, but allows for proof-of-mechanism studies of targeted therapeu tic agents active against the genetic lesions produced
in such animals (Figure 2a). This exemplifies how the exploration of human tumour biology in model systems has been enhanced through the use of GEMs. Recent efforts to improve the clinical use of GEMs, which have been hindered by the prolonged timelines (often 9–12 months) and expense required to generate the models, have focused on the possibility that orthotopic or subcutaneous transplantation of tumours initially established in GEMs could facilitate their use for drug screening (Figure 2b). GEM allografts may enhance the application of clinical trial design principles in ‘preclini cal clinical trials’ so that these models could more easily be employed with adequate sample sizes, predefined therapeutic hypotheses, and integrated pharmaco dynamic and pharmacokinetic studies, facilitating comparisons of drug efficacy in tumours from GEMs
NATURE REVIEWS | CLINICAL ONCOLOGY
ADVANCE ONLINE PUBLICATION | 3 © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS with human clinical trial results of the same agents and combinations.34 Additional progress will expand the diversity of GEM models to characterize more fully the wide range of molecular aberrations observed in clinical tumour specimens; there is also little question that greater use of GEM technology could assist efforts focusing on biomarker development for early stage disease as well as the evolution of therapeutic resistance. There has also been a dramatic resurgence of interest in mouse tumour models generated directly from patient surgical or biopsy samples, denoted as patient-derived tumour xenografts (PDXs). 36 Although such model systems had been investigated for decades, the availabil ity of NOD/SCID and NOD/SCID/IL2Rγnull (NSG) mice over the past 10 years, which enhances murine immunoincompetence by blocking natural killer cell maturation and facilitates tumour engraftment, has led to a renais sance in the use of this technique for generating tumour models. Furthermore, for certain solid tumours, such as colorectal, pancreatic, lung, and breast cancers, early tumour passages (probably through the third or fourth generation in most cases) retain a degree of human stroma, facilitating studies of gene expression, drug effi cacy, and tumour heterogeneity.36 In the case of endo crine sensitive and resistant breast cancer, PDX models have been used to demonstrate the stability of genomewide allele frequencies, and response to endocrine therapy, between the initial human tumour specimens and PDXs established from these tissues.37 Therefore, the development of large panels of clinically-annotated tumours could help evaluate the diversity of responses expected in the clinical arena, or could be used for molecularly-characterized preclinical trials to assess the validity of target selection strategies in advance of clini cal testing (Figure 2c). These models avoid the problems associated with in vitro tumour cell selection; however, the absence of an intact immune system negates their value for testing immunotherapeutic strategies. Most PDX models have been developed from surgical speci mens rather than from biopsies of metastatic sites; thus, these models might not be representative of the tumours that occur in patients with recurrent malignancies who require systemic therapy.38 Further development of PDX models for both biological studies and drug selection might have an important role in improving the precision with which cancer treatments are selected in the future. While there has been considerable focus on improv ing in vivo models of human cancer, recent studies have suggested better ways to study human tumours in vitro. The immortalization of epithelial cells, including human tumour cells, by inhibiting terminal differentiation through the use of an inhibitor of Rho kinase and con ditioned media, has excited investigational interest; if widely applicable, this ‘conditional reprogramming’ tech nique might facilitate the long-term culture of tumour cells that heretofore have been difficult to propagate, such as tumours of the prostate (Figure 2c).39,40 It is not known whether tumour cells cultured in this fashion undergo genetic adaptation to growth on plastic, which would diminish their predictive potential. The development of
3D cultures of cancer organoids—tumour cells that grow in a collagen-rich matrix supported by essential growth factors and that have phenotypic characteristics of cancer stem cells—is the other major focus for the development of in vitro cancer models. Although these models lack stroma and require very well-defined growth conditions, organoids can be propagated indefinitely from indivi dual patients, are amenable to drug screening, and can retain essential molecular characteristics of the primary disease (at least for colon cancer).41,42 There is great inter est in the possibility that organoid cultures developed for other malignancies (such as pancreatic and prostate cancers) will provide important new in vitro tools for cancer biology and drug discovery.41,42 Inflammation and cancer One of the most important translational research relationships to understand is the interaction between innate inflammatory responses, host defence, and carcinogenesis; our understanding of this relationship, including the role of the human microbiome in these pro cesses, has expanded dramatically over the past decade.43 A deeper appreciation of the molecular mechanisms underlying chronic inflammation-related cancers (such as those associated with inflammatory bowel disease, chronic pancreatitis, and viral hepatitis) and the evidence for the protective role of non-steroidal anti-inflammatory agents in the prevention of multiple malignant histolo gies, indicates that greater emphasis on the development of new therapeutic modalities to prevent the adverse con sequences of the unchecked cytokine release and reactive oxygen production that accompany chronic inflamma tory stress are urgently required. Such therapeutics might interdict the proinflammatory microenvironment that is conducive to DNA damage, cancer initiation, and pro gression.44–47 Recent evidence has demonstrated previ ously unknown relationships between the inflammatory environment, produced by commensal bacteria, and the efficacy of both immune and cytotoxic anticancer thera pies in vivo, emphasizing the role of the tumour micro environment in the control of therapeutic response.48 It seems likely that in the near future a better understand ing of how to control both proinflammatory and antiinflammatory responses will play a critical part in the developmental therapeutics of cancer. Cancer metabolism Another remarkable change in translational cancer biology over the past decade has been the recognition that the metabolic reprogramming of tumour cells is inti mately related to oncogene-induced proliferative signal ling and the function of tumour suppressor genes.49,50 Thus, signal transduction pathways that are important for malignant transformation (for example, those controlled by Myc, p53, and mTOR) have a critical role in altering the metabolic phenotypes of cancer cells, while at the same time the products of altered tumour cell metabolism (α-ketoglutarate and reactive oxygen species [ROS]) affect tumour cell signal transduction.51 To enhance cell pro liferation, tumour cells adapt their metabolic machinery
4 | ADVANCE ONLINE PUBLICATION
www.nature.com/nrclinonc © 2014 Macmillan Publishers Limited. All rights reserved
REVIEWS to use a variety of energy sources besides glucose (such as glutamine) to increase the production of fatty acids and other macromolecules needed for the production of cell membranes as well as a diverse portfolio of anabolic processes, in addition to ATP production. The metabolic programmes of tumour cells are diverse, tissue context dependent, highly influenced by the surrounding micro environment as well as the tumour host, and hetero geneous even within a single tumour.52 Recent evidence suggests that metabolic reprogramming of tumour cells might have a critical role in the epigenetic modification of tumour cell DNA and histones.53 For these reasons, substantial investigational effort is now being directed toward the development of small-molecule inhibitors of specific metabolic enzymes that contribute to tumour cell proliferation (such as isocitrate dehydrogenase 1 [IDH1] and IDH2) for low-grade gliomas and acute myelogenous leukaemia, respectively.54,55
Characterization of tumours and host tissues About the same time that imatinib was found to be an effective treatment for chronic myelogenous leukaemia and erlotinib for adenocarcinoma of the lung with EGFR mutations, the sequencing of the human genome was completed.56 The completion of this project propelled the application of genomic sequencing technologies in cancer biology (Figure 1). Although much effort has been focused on the presence of somatic mutations in specific cancer histologies,57 the availability of powerful sequencing techniques also led to a series of populationbased studies to discover germline mutations that could confer cancer risk, so-called genome-wide association studies.58 To date over 150 regions of the human genome, often in noncoding domains, have been correlated with two dozen human cancers. A major area of current investigation is to understand how these genetic vari ants predispose to cancer, and whether or how inter actions occur between somatic and germline variants in cancer development.59 Human biospecimens Molecular tumour characterization, whether it invol ves DNA sequencing, evaluation of RNA expression, proteomic or phosphoproteomic determinations, or immunohistochemistry, requires the collection of high quality biospecimens.60 The lack of accepted national guidelines for specimen acquisition stimulated a multiyear process leading to the introduction of standards for the collection and banking of both tumour and normal tissues.61 Lack of standard biospecimen collec tion, processing, and storage procedures hampered the initiation of the TCGA project; difficulties were encoun tered in obtaining tumour specimens that met quality standards for sequencing in terms of the tumour cell purity of the samples; the fixation method employed; and available volumes of high-quality tumours across several histologies. The value of proof-of-mechanism pharmacodynamic assays is frequently diminished because insufficient care is taken to understand the stability of the analyte of interest
in the context of performing research tumour biopsies; sample handling techniques that will stabilize biomarkers after tissues are acquired are infrequently evaluated before study initiation.62–64 Unfortunately, regardless of the bio marker, there is a need to improve the details of sample acquisition that are fundamental to successful molecular characterization of a tumour or normal tissue of interest. Clinical characterization of tumours The availability of massively parallel DNA sequencing in the mid‑2000s allowed for an exponential increase in the speed and amount of tumour DNA sequence that could be obtained in a given period of time; this enhanced efficiency led to plummeting costs for acquiring DNA sequences and the development of better tools to analyse sequence data.28 Systematic DNA sequencing substan tiated the role of receptor tyrosine kinase mutations in driving tumour cell signalling across many cancers (PI3KCA in breast cancer and JAK2 in myeloprolifera tive syndromes) and has been essential for stimulating the development of new kinase inhibitors for the treat ment of tumours that are dependent on such pathways. The mutational patterns observed—extremely high fre quencies in lung cancer and melanoma, and relatively low levels in untreated paediatric malignancies—have also substantiated aetiological associations with the extent and duration of mutagen exposure in the former malignancies, and have suggested hypotheses for the development of tumours in non-renewing tissue types for the latter.65 A recent analysis indicates that despite the intensive investigation of >10,000 tumours by the TCGA, additional novel driver mutations are likely to be found at low frequency (
