Personalised cancer care: promises and challenges of targeted therapy

Abstract

Personalised cancer care has great appeal for many reasons. Because it presumes that one tailors specific treatments for the right cancers, personalised care implies that doctors will be providing more effective treatments with less toxic effects. Because it predicates that one identifies the various subtypes of cancer, personalised care suggests that scientists will have resolved the complexity of cancer.

A fundamental premise of personalised care is that cancer subtypes can be discerned using unique genetic and molecular biomarkers or signatures. Considering that epigenetics and the microenvironment also play an essential role in the final manifestation of the malignant process, we propose that a more comprehensive or unified theory of cancer is needed to attain personalised cancer care, in which one targets not only the genetic components but also the entire cellular make-up of individual cancer subtypes.

Personalised care: a brief history of cancer cures

In many respects, personalised care is already being practised to treat several curable cancers, such as testicular non-seminomatous germ cell tumour, acute promyelocytic leukaemia and diffuse large B-cell lymphoma. Because one can clinically distinguish these malignant subtypes, it is feasible to individualise care and assign specific treatments.

Friedman and Moore¹ classified four different subtypes of testicular germ cell tumours. Since 1980, chemotherapy has revolutionised the treatment of advanced germ cell tumours;² however, surgery is still indispensable for the treatment of teratomas, and both chemotherapy and surgery are required to cure mixed germ cell tumours containing teratomas.

The French–American–British classification system categorised six subtypes of acute myelocytic leukaemia.³ Since 1991, all trans-retinoic acid and arsenic trioxide have transformed the treatment of acute promyelocytic leukaemia.^4,5 Although more than 80% of patients with acute promyelocytic leukaemia experience a complete remission with all trans-retinoic acid or arsenic trioxide, fewer than 10% will do if all acute myelocytic leukaemias have been treated as one disease with the same treatment.

The Working formulation first divided non-Hodgkin lymphoma into four subtypes.⁶ Subsequently, the Revised European–American Lymphoma Classification⁷ used morphology, immunophenotype, genetic characteristics and clinical features to distinguish 14 distinct non-Hodgkin B-cell lymphomas. Today, chemotherapy cures about 60% of patients with advanced diffuse large B-cell lymphoma.⁸

In 2001, the Human Genome Project was launched.^9,10 An unspoken goal of this project was personalised medicine: When one could treat a certain cancer subtype on the basis of cancer genetics, personalised care would replace empiric care. No longer would doctors prescribe treatments by trial-and-errors or hit-and-misses. It is time to forsake the concept of a ‘one-treatment-fits-all’ approach.

Unlike testicular cancer, acute leukaemia or aggressive lymphoma, not many malignant tumours have been cured in a substantive way since 2001. We need a strategy to move above and beyond the conventional thinking of cancer genetics in an effort to develop curative targeted therapies and personalised care.

Cancer genetics

When one considers key genetic targets in cancer, a pattern emerges. In general, oncology recapitulates ontogeny. Interestingly, an oncogene is actually a defective proto-oncogene. And a malignant cell has all the characteristics of an aberrant stem cell.^11–14 Hence, many key genetic targets in cancer involve stemness pathways, including pluripotency, differentiation, self-renewal and asymmetric division. Importantly, when it concerns stemness pathways, intercellular interactions and the microenvironment become integral factors in a complete equation of cancer. This interconnectedness constitutes a vital tenet of the unified theory of cancer.

For example, tyrosine kinases (e.g. Epidermal growth factor, Fibroblast growth factor and Vascular endothelial growth factor) are better known as growth factors in adult organisms, but they also play critical roles as morphogens during development.¹⁵ The tyrosine kinase families are evolutionarily conserved from fruit fly to man and serve as master regulators of position-dependent cell fate during embryogenesis. Hence, genetic targets involved in morphogenesis are also engaged in morphostasis. The importance of morphogenetic pathways, such as RAS-RAF-MAP kinase, Phosphoinositide 3-kinase and AKT, for the maintenance of epithelial homeostasis is underscored by the fact that mutations disrupting them contribute to carcinogenesis.

A compilation of representative genetic targets and their associated targeted therapies is listed in Table 1. It is evident that many genetic targets overlap (e.g. human epidermal growth factor receptor 2 involves both stemness and mitogenic pathways), and several targeted therapies have multiple genetic targets. One cannot deny the success of pioneering treatments, such as imatinib for chronic-phase bcr-abl Chronic myelogenous leukemia,¹⁶ trastuzumab for human epidermal growth factor receptor 2-positive breast cancer,¹⁷ vemurafenib for BRAF^V600E melanoma¹⁸ and erlotinib for Epidermal growth factor receptor mutant-mutant non-small cell lung cancer.¹⁹ But, something is amiss. We have not yet cured any cancers with targeted therapy, and the duration of response tends to be transient or limited.^20,21 Invariably, drug resistance occurs and additional targeted therapies need to be devised.

Table 1.

Genetic targets and targeted therapies.

Cancer hallmark	Genetic target	Targeted therapy	Cancer type	FDA approval
Stemness	HER-2	Trastuzumab	Breast cancer	1998
Stemness	c-KIT	Imatinib	Gastrointestinal stromal tumours	2002
Stemness	JAK-1, JAK-2	Ruxolitinib	Myelofibrosis	2011
Stemness	EML4-ALK	Crizotinib	Non-small cell lung cancer (NSCLC)	2011
Stemness	SMO	Vismodegib	Basal cell cancer	2012
Differentiation	ER	Tamoxifen	Breast cancer	1977
Differentiation	PML-RAR	All-trans retinoic acid	Acute promyelocytic leukaemia	1995
Differentiation	CD20	Rituximab	B-cell lymphomas	1997
Differentiation	PML-RAR	Arsenic trioxide	Acute promyelocytic leukaemia	2000
Differentiation	Dnmt	Azacitabine	Myelodysplastic syndrome	2004
Differentiation	HDAC	Vorinostat	Cutaneous T cell lymphoma
Differentiation	CYP17A1	Abiraterone	Prostate cancer	2011
Differentiation	AR	Enzalutamide	Prostate cancer	2012
Angiogenesis	VEGF-A	Bevacizumab	Colorectal cancer NSCLC (non-squamous) Renal cell cancer Glioblastoma	2004 2006 2009 2009
Mitogenic	Bcr-abl	Imatinib	Chronic myelogenous leukaemia	2001
Mitogenic	EGFR	Erlotinib	NSCLC Pancreatic cancer	2004 2005
Mitogenic	VEGFR PDGFR C-Raf	Sorafenib	Renal cell cancer Hepatocellular cancer	2005 2007
Mitogenic	EGFR	Cetuximab	Head and neck cancer	2006
Mitogenic	mTOR	Everolimus	Renal cell cancer	2009
Mitogenic	BRAF V600E	Vemurafenib	Melanoma	2011
Mitogenic	VEGFR EGFR c-RET	Vandetanib	Medullary thyroid cancer	2011
Mitogenic	MEK1, MEK2	Trametinib	Melanoma	2013
Immunity	CTLA-4	Ipilimumab	Melanoma	2011
Immunity	PD-1L	Pembrolizumab	Melanoma	2014

FDA: Food and Drug Administration.

Targeted therapy

The idea that cancer subtypes are better defined by their cells of origin than by their specific genetic defects is supported by clinical evidence showing a paucity of reliable biomarkers and curative therapies when the targets are based on genetic defects rather than cellular origins. Perhaps, one needs to target a whole network of genetic signatures rather than a few selected genetic defects. But when one deals with a whole network of genetic signatures, a distinct cellular entity with a unique cellular origin becomes manifested.

Considering that p53 is the most commonly mutated cancer gene in humans, one would expect p53-targeted therapy to be a boon to cancer therapy. However, the results of a phase III study did not support the prognostic value of p53 or the benefit of adjuvant methotrexate, vinblastine, doxorubicin, cisplatin chemotherapy in patients with p53-positive bladder cancer.²² A pilot study using ONYX-015 for the treatment of oral dysplasia showed that 37% of lesions underwent histologic regression, but the responses were short-lived.²³ In addition, RAS pathway mutations occur in >90% of pancreatic cancers. However, cetuximab and gemcitabine did not improve the survival duration of patients with metastatic pancreatic cancer compared with gemcitabine alone.²⁴ Thus, a disconnect seems to exist between targeted therapy and its most heralded targets.

Perhaps an alternative unified theory of cancer is needed to account for the many contrary findings. For example, patients with human epidermal growth factor receptor 2-positive tumours benefited from adjuvant trastuzumab, but so did those with human epidermal growth factor receptor 2-negative tumours.²⁵ Such finding contradicts the very essence of targeted therapy or personalised care. Indeed, there is evidence linking human epidermal growth factor receptor 2 to cellular stemness, suggesting that it plays a central role in the intrinsic regulatory pathways of breast cancer stem cells and their interaction with the tumour microenvironment.²⁶ Similarly, olaparib (a poly ADP-ribose polymerase inhibitor) resulted in a 41% objective response rate in patients with BRCA1 or BRCA2 mutations but also a 24% rate in those without the mutations.²⁷ Thus, it seems that targeted therapy is sometimes effective regardless of whether its putative target is present or not.

One would expect tumours with plentiful mutations to be generally more lethal than those with few mutations and tumours with certain mutations to be potentially more threatening than those without them. Paradoxically, more aggressive tumour subtypes may contain fewer mutations than their less aggressive counterparts.^28–30 Furthermore, the prognosis of patients with colon cancer and ‘microsatellite instability’ tumours was found to be better than that of patients with microsatellite-stable tumours.³¹ And patients with advanced Epidermal growth factor receptor-overexpressing non-small cell lung cancer had longer overall survival durations than those with normal Epidermal growth factor receptor-functioning tumours.^32,33 In the case of ependymonas, there are actually no gene mutations that could account for the presence or cause of a malignancy or that could be used as a prognostic biomarker or targeted therapy for personalised cancer care.³⁴

Promises and challenges

An inconvenient truth about targeted therapy is that the results have not been commensurate with expectations. Although the promises of targeted therapy are self-evident, its challenges are not as obvious. Perhaps, any tangible clinical benefits derived from personalised cancer care have been modest or marginal, because the fundamental concepts of targeted therapy are not aligned with a unified theory of cancer. To fulfill the promises and meet the challenges of targeted therapy and personalised cancer care, one needs to return to the basics of cancer biology. As a starter, one needs to unravel the enigma of driver mutations, intratumoural heterogeneity and cancer subtypes.

Driver mutations

When it is necessary to pick among a plethora of genetic targets, a reliable method is needed to prioritise which ones have the best chance of benefiting patients in a meaningful way, i.e. in a way that is clinically significant, not just statistically significant. Nowadays, one realises that it is not feasible to interrogate all genetic targets, many of which are incidental rather than causal, and for which treatments may prove to be futile or even harmful. It is imperative that one devises a method to distinguish the so-called driver from passenger mutations, and discerns those genetic targets that are true makers rather than mere markers of the malignant processes.

The scientific method (i.e. pertinent observation, useful hypothesis, informative experiments and empirical validations)³⁵ may enable one to determine whether a genetic target is the cause or an effect of cancer. Paradoxically, the current modus operandi of targeted therapy seems to be intrinsically incongruous with the fundamental principles of the scientific method. First of all, all pertinent observations indicate that cancer is a complex disease rather than a simple entity. In addition, many useful theories suggest that cancer is a multifaceted problem (e.g. genetics, epigenetics and microenvironment) in need of an integrated solution. Furthermore, informative experiments repeatedly demonstrate that several important cancer hallmarks, including metastasis, heterogeneity and dormancy, do not actually depend on any specific genetic mutations.^36–41 In light of this irony, targeted therapy seems to be an oversimplification and driver mutations may be a myth.

Another method to determine whether a genetic target is the cause or an effect of cancer is to perform multivariant analysis⁴² of composite genetic targets. Although targeted therapy is the product of a reductionist approach based on the traditional method of dissecting the whole and examining the parts in the laboratory, one cannot help but observe that cancer is inherently heterogeneous, interactive and dynamic in the clinic. Whereas a particular genetic target may be the ‘driver’ in isolation, it could be but one of the many ‘passengers’ in toto. By considering the whole system and its web of intricate networks, one could apply multivariant analysis to discover the real-driver mutations and design breakthrough rather than incremental targeted therapies for the benefit of personalised cancer care.

Intratumoural heterogeneity

Elucidating intratumoural heterogeneity is tantamount to overcome the challenges of targeted therapy and personalised cancer care. If intratumoural heterogeneity arises from the action of defective genes, then one targets and fixes them. On the other hand, if intratumoural heterogeneity is determined by unique cellular (e.g. stem-like) entities in which these genetic changes occur, then one needs to treat both the aberrant pluripotent progenitor cells and the diverse differentiated cancer cells.

For example, a classic targeted therapy that many people forget is androgen deprivation for the treatment of prostate cancer.⁴³ Almost all patients, including those with advanced metastatic disease, respond to the treatment often for years, sometimes for decades. However, androgen deprivation therapy is not a curative treatment, even though it is effective. To improve on this prototype therapy that targets differentiated prostate cancer cells, one needs to integrate it with treatments that affect other tumour compartments (e.g. prostate cancer stem cells) and the ubiquitous tumour microenvironment (e.g. bone stromal cells).

Testicular cancer provides another example of intratumoural heterogeneity in solid tumours. Because of clonal origin, it is well known that the different histologic components of germ cell tumour have a similar, if not the same, genetic profile.^44–46 Hence, both chemo-sensitive embryonal carcinoma and chemo-resistant teratoma in a mixed germ cell tumour have the same wild-type p53 or a defective B-cell lymphoma 2.^47,48 When a genetic mutation is observed in both the cancer stem cell and the differentiated cancer cell within the same tumour, and yet the disparate cells have distinct biological and clinical characteristics, it seems fitting that targeting the disparate cells rather than the mutation itself would be more efficacious.^49,50 Indeed, by treating the distinct cell types rather than the specific genetic mutations, one has rendered testicular cancer the most curable solid tumours known to date.

Recently, several investigators showed that intratumour heterogeneity can compromise the value of biomarkers, as determined by using single tumour biopsy samples.^51–53 The results suggested that genetic approach may be inadequate for solving the problem of tumour heterogeneity and for reaching our goal of personalised cancer care. Kreso et al.⁵⁴ also demonstrated that individual cancer cells within a uniform genetic lineage were functionally heterogeneous with respect to their growth dynamics, persistence through serial transplantation and response to therapy. Thus, non-genetic processes (altered epigenetic states and microenvironmental influences) also generate heterogeneity.

Therefore, personalised cancer care may be more effective when aberrant cells are targeted rather than defective genes. This approach implies that one should target the microenvironment of aberrant cells and the aberrant cells themselves, as they contain tens, hundreds or even thousands of relevant and irrelevant defective genes.^55,56 When one targets cancer cells at their core, namely their cell of origin, both cancer stem cells and differentiated cancer cells are being accounted for. When one treats the intracellular, intercellular and microenvironmental factors all at once, the presence of incidental or redundant genetic defects becomes less of a concern.

Cancer subtypes

Solving intertumoural heterogeneity is another prerequisite to fulfilling the promises of targeted therapy and personalised cancer care. One way to accomplish this task is to identify cancer subtypes. By identifying subtypes, one renders a complex disease into simple ones, which can then be targeted. A popular strategy for identifying cancer subtypes is to use biomarkers. An alternative is to decipher their cells of origin. Recent evidence suggests that the key to unlocking the secrets of intertumoural heterogeneity may be found in the cellular profiles rather than the molecular genetics of cancer.

The Cancer Genome Atlas Network⁵⁷ used six different technology platforms (genomic deoxyribonucleic acid copy number arrays, deoxyribonucleic acid methylation, exome sequencing, messenger ribonucleic acid arrays, micro-ribonucleic acid sequencing and reverse-phase protein arrays) to analyse the genes of 825 breast cancer patients and confirmed the existence of four major subtypes of breast cancer: luminal A, luminal B, basal-like (triple-negative) and human epidermal growth factor receptor 2 enriched. Clinically observable plasticity and heterogeneity occur within but not across these major biological subtypes of breast cancer, suggesting that they have distinct cellular origins. Although there is enrichment of specific mutations within the tumour subtypes (i.e. phosphatidylinositol-4,5-bisphosphate 3-kinase [49%] and Mitogen-Activated Protein Kinase Kinase Kinase 1 [14%] in luminal A and TP53 [84%] in basal-like), it is uncommon for individual tumours to harbour the same genetic mutations even within the same subtype.⁵⁷ Interestingly, only about 50% of clinically human epidermal growth factor receptor 2 positive tumours fall into the human epidermal growth factor receptor 2 enriched-messenger ribonucleic acid subtype/human epidermal growth factor receptor 2-protein group, while the rest of the clinically human epidermal growth factor receptor 2 positive tumours is predominantly observed in the luminal-messenger ribonucleic acid subtypes.⁵⁷

Medulloblastoma can be parsed into unique subtypes with associated risk stratifications: Wnt pathway-active, excellent prognosis; Hedgehog pathway-active, good prognosis; and c-MYC amplification, poor prognosis.⁵⁸ Since all subtypes (i.e. different diseases) showed elevated MYC expression (i.e. the same target), it seems that the cell of origin, rather than a particular molecular defect, is paramount in the manifestation of a malignant phenotype. Indeed, Gibson et al.⁵⁹ showed that the subtypes of medulloblastoma have distinct developmental origins: the Wnt subtype arises from the lower rhombic lip and embryonic dorsal brainstem, outside the cerebellum, whereas the Sonic Hedgehog subtype originates from the upper rhombic lip and developing cerebellum.

Brannon et al.⁶⁰ surmised that the heterogeneity of clear cell renal cell carcinoma could be traced to multistep carcinogenesis, i.e. acquisition of the Von Hippel-Lindau mutation, Hypoxia-inducible factor stabilisation and secondary mutations. However, the identification of two unique renal cell carcinoma subtypes with distinct survival patterns suggests that ccB arose from earlier progenitor cells (overexpression of genes associated with Epithelial -mesenchymal transition, Transforming growth factor-β, Wnt and wound healing), whereas ccA originated from later progenitor cells (association with genes involved in angiogenesis and fatty acid metabolism).⁶⁰ Importantly, the Von Hippel-Lindau mutation itself did not differentiate between the two subtypes.

There are two distinct subtypes of posterior fossa ependymona: posterior fossa group B occurs in older children and adults and carries a very good prognosis, whereas posterior fossa group A is mainly seen in infants and is associated with poor prognosis despite intensive therapies. Different radial glial cells from different regions of the nervous system have been shown to be the cells of origin for the different subtypes of ependymonas.⁶¹ Importantly, whole genome and whole exome sequencing across a broad cohort (n = 47) of posterior fossa ependymonas did not detect any significant or recurrent gene mutations that could account for the disparate subtypes of ependymona.³⁴

Conclusions

Although there is great promise and general acceptance in the use of cancer genomics for personalised care, we still need to reconcile the genetic with a cellular origin of cancer. We propose a unified theory of cancer to resolve the dilemma of targeted therapy for personalised cancer care. We advocate an integrated approach that considers both the genetic and a cellular origin of cancer to fulfill the promises and meet the challenges of targeted therapy and personalised cancer care.

Footnotes

Declarations

References

Friedman

Moore

. Tumors of the testis: a report on 922 cases. Mil Surg 1946; 99: 573–593.

Bosl

Gluckman

Geller

. An effective chemotherapy regimen for patients with germ cell tumors. J Clin Oncol 1986; 4: 1493–1499.

Bennett

Catovsky

Daniel

. proposals for the classification of the acute leukaemias. French-American-British (FAB) co-operative group. Br J Haematol 1976; 33: 451–458.

Warrell

Jr Frankel

Miller

WHJ

. Differentiation therapy of acute promyelocytic leukemia with tretinoin (all-trans-retinoic acid). N Engl J Med 1991; 324: 1385–1393.

Soignet

Frankel

Douer

. United States multicenter study of arsenic trioxide in relapsed acute promyelocytic leukemia. J Clin Oncol 2001; 19: 3852–3860.

The Non-Hodgkin’s Lymphoma Pathologic Classification Project. National Cancer Institute sponsored study of classifications of non-Hodgkin’s lymphomas: summary and description of a Working Formulation for clinical usage. Cancer 1982; 49: 2112–2135.

Harris

Jaffe

Stein

. A revised European-American classification of lymphoid neoplasms: a proposal from the International Lymphoma Study Group. Blood 1994; 84: 1361–1392.

Coiffier

Thieblemont

Van den Neste

. Long term outcome of patients in the LNH-98.5 trial, the first randomized study comparing rituximab-CHOP to standard CHOP chemotherapy in DLBCL patients: a study by the Groupe d’Etudes des Lymphomes de L’Adultes. Blood 2010; 116: 2040–2045.

Lander

Linton

Birren

. Initial sequencing and analysis of the human genome. Nature 2001; 409: 860–921.

10.

Venter

Adams

Myers

. The sequence of the human genome. Science 2001; 291: 1304–1351.

11.

Origin of cancers. Clinical perspectives and implications of a stem-cell theory of cancer. In: Rosen

(ed). Cancer Treatment and Research 2010; Vol. 154, New York, NY: Springer, pp. 57–57.

12.

Lin

Logothetis

. Stem-cell origin of metastasis and heterogeneity in solid tumours. Lancet Oncol 2002; 3: 508–513.

13.

. Cancer: a “stem-cell” disease? Cancer Cell Int 2013; 13: 40–40.

14.

Som

Wen

. Stem cell origin of testicular seminoma. Clin Genitourin Cancer 2013; 11: 489–494.

15.

Van den Brink

Offerhaus

. The morphogenetic code and colon cancer development. Cancer Cell 2007; 11: 109–117.

16.

Druker

Talpaz

Resta

. Efficacy and safety of a specific inhibitor of the BCR-ABL tyrosine kinase in chronic myeloid leukemia. N Engl J Med 2001; 344: 1031–1037.

17.

Piccart-Gebhart

Procter

Leyland-Jones

. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005; 353: 1659–1672.

18.

Chapman

Hauschild

Haanen

. Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N Engl J Med 2011; 364: 2507–2516.

19.

Pirker

Pereira

Szczesna

. Cetuximab plus chemotherapy in patients with advanced non-small-cell lung cancer (FLEX): an open-label randomized phase III trial. Lancet 2009; 373: 1525–1531.

20.

Gore

Larkin

. Challenges and opportunities for converting renal cell carcinoma into a chronic disease with targeted therapies. Br J Cancer 2011; 104: 399–406.

21.

Diaz

Jr Williams

. The molecular evolution of acquired resistance to targeted EGFR blockade in colorectal cancers. Nature 2012; 486: 537–540.

22.

Stadler

Lerner

Groshen

. Phase III study of molecularly targeted adjuvant therapy in locally advanced urothelial cancer of the bladder based on p53 status. J Clin Oncol 2011; 29: 3443–3449.

23.

Rudin

Cohen

Papadimitrakopoulou

. An attenuated adenovirus, ONYX-015, as mouthwash therapy for premalignant oral dysplasia. J Clin Oncol 2003; 21: 4546–4552.

24.

Philip

Benedettii

Corless

. Phase III study comparing gemcitabine plus cetuximab versus gemcitabine in patients with advanced pancreatic adenocarcinoma: Southwest Oncology group-directed intergroup trial S0205. J Clin Oncol 2010; 28: 3605–3610.

25.

Paik

Kim

Wolmark

. HER2 status and benefit from adjuvant trastuzumab in breast cancer. N Engl J Med 2008; 358: 1409–1411.

26.

Korkaya

Wicha

. Breast cancer stem cells: we have got them surrounded. Clin Cancer Res 2013; 19: 511–513.

27.

Gelmon

Tischkowitz

Mackay

. Olaparib in patients with recurrent high-grade serous or poorly differentiated ovarian carcinoma or triple-negative breast cancer: a phase 2, multicenter, open-label, non-randomised study. Lancet Oncol 2011; 12: 852–861.

28.

Kuppers

Klein

Hansmann

Rajewsky

. Cellular origin of human B-cell lymphomas. N Engl J Med 1999; 341: 1520–1259.

29.

Bergamaschi

Kim

Wang

. Distinct patterns of DNA copy number alterations are associated with different clinicopathological features and gene-expression subtypes of breast cancer. Genes, Chromosomes Cancer 2006; 45: 1033–1040.

30.

Walter

Shen

Ding

. Clonal architecture of secondary acute myeloid leukemia. N Engl J Med 2012; 366: 1090–1098.

31.

Ribic

Sargent

Moore

. Tumor microsatellite-instability status as a predictor of benefit from fluorouracil-based adjuvant chemotherapy for colon cancer. N Engl J Med 2003; 349: 247–257.

32.

Tsao

Sakurada

Cutz . Erlotinib in lung cancer – molecular and clinical predictors of outcome. N Engl J Med 2005; 353: 133–144.

33.

Hirsch

Varella-Garcia

Bunn

Jr . Molecular predictors of outcome with gefitinib in a phase III placebo-controlled study in advanced non-small cell lung cancer. J Clin Oncol 2006; 24: 5034–5042.

34.

Mack

Witt

Piro

. Epigenomic alterations define lethal CIMP-positive ependymomas of infancy. Nature 2014; 506: 445–450.

35.

Bilen

Tannir

. The scientific method: pillar and pitfall of cancer research. Cancer Med 2014; 3: 1035–1037.

36.

Walther

Johnstone

Swanton

. Genetic prognostic and predictive markers in colorectal cancer. Nat Rev Cancer 2009; 9: 489–499.

37.

Jones

Chen

Parmigiani

. Comparative lesion sequencing provides insights into tumor evolution. Proc Natl Acad Sci USA 2008; 105: 4283–4288.

38.

Cheng

Liu

Yoshida

. Lineage infidelity of epithelial cancers is controlled by HOX genes that specify regional identity in the reproductive tract. Nat Med 2005; 11: 531–537.

39.

Hingorani

Wang

Multani

. Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice. Cancer Cell 2005; 7: 469–483.

40.

Ince

Richardson

Bell

. Transformation of different human breast epithelial cell types leads to distinct tumor phenotypes. Cancer Cell 2007; 12: 160–170.

41.

Michor

Iwasa

Nowak

. Dynamics of cancer progression. Nat Rev Cancer 2004; 4: 197–205.

42.

Anderson

. An introduction to multivariate statistical analysis, 3rd ed. New York, NY: John Wiley and Sons, 2004.

43.

Huggins

Hodges

. Studies on prostatic cancer: effects of castration, of estrogen, and of androgen injection on serum phosphatases in metastatic carcinoma of the prostate. Cancer Res 1941; 1: 293–297.

44.

Kernek

Ulbright

Zhang

. Identical allelic losses in mature teratoma and other histologic components of malignant mixed germ cell tumors of the testis. Am J Pathol 2003; 163: 2477–2484.

45.

Jones

Wang

Sung

. Clonal origin of metastatic testicular teratomas. Clin Cancer Res 2006; 12: 5377–5383.

46.

Kum

Ulbright

Williamson

. Molecular genetic evidence supporting the origin of somatic-type malignancy and teratoma from the same progenitor cell. Am J Surg Pathol 2012; 36: 1849–1856.

47.

Burger

Nooter

Boersma

. Lack of correlation between cisplatin-induced apoptosis, p53 status, and expression of Bcl-2 family proteins in testicular germ cell tumour cell lines. Int J Cancer 1997; 73: 592–599.

48.

Cavallo

Feldman

Barchi

. Revisiting DNA damage, p53-mediated apoptosis, and cisplatin sensitivity in germ cell tumors. Int J Dev Biol 2013; 57: 273–280.

49.

Cheng

Liu

Yoshida

. Lineage infidelity of epithelial cancers is controlled by HOX genes that specify regional identity in the reproductive tract. Nat Med 2005; 11: 531–537.

50.

Hingorani

Wang

Multani

. Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice. Cancer Cell 2005; 7: 469–483.

51.

Navin

Krasnitz

Rodgers

. Inferring tumor progression from genomic heterogeneity. Genome Res 2010; 20: 68–80.

52.

Nickel

Barnholtz-Sloan

Gould

. Characterizing mutational heterogeneity in a glioblastoma patient with double recurrence. PLoS One 2012; 7: e35262–e35262.

53.

Gerlinger

Rowan

Horswell

. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 2012; 366: 883–892.

54.

Kreso

O’Brien

van Galen

. Variable clonal repopulation dynamics influence chemotherapy response in colorectal cancer. Science 2012; 339: 543–548.

55.

Sjoblom

Jones

Wood

. The consensus coding sequences of human breast and colorectal cancers. Science 2006; 314: 268–274.

56.

Berger

Lawrence

Demichelis

. The genomic complexity of primary human prostate cancer. Nature 2011; 470: 214–220.

57.

The Cancer Genome Atlas Network. Comprehensive molecular portraits of human breast tumours. Nature 2012; 490: 61–70.

58.

Monje

Beachy

Fisher

. Hedgehog, flies, Wnts, and MYCs: the time has come for many things in medulloblastoma. J Clin Oncol 2011; 29: 1395–1398.

59.

Gibson

Tong

Robinson

. Subtypes of medulloblastoma have distinct developmental origins. Nature 2010; 468: 1095–1099.

60.

Brannon

Reddy

Seiler

. Molecular stratification of clear cell carcinoma by consensus clustering reveals distinct subtypes and survival patterns. Genes Cancer 2010; 1: 152–163.

61.

Taylor

Poppleton

Fuller

. Radial glia cells are candidate stem cells of ependymona. Cancer Cell 2005; 8: 323–335.