Histopathology 2019, 75, 4–19. DOI: 10.1111/his.13828 REVIEW Integrating clinical, molecular, proteomic and histopathological data within the tissue context: tissunomics Santiago Ram on y Cajal,1,2,3 Stefan H€ummer,1,3 Vicente Peg,2,3 Xavier M Guiu,3,4 2 2,3 Ines De Torres, Josep Castellvi, Elena Martinez-Saez2 & Javier Hernandez-Losa1,2,3 1 Translational Molecular Pathology, Vall d’Hebron Institute of Research (VHIR), Universitat Autonoma de Barcelona, 2 Department of Pathology, Vall d’Hebron University Hospital, 3Spanish Biomedical Research Network Centre in Oncology (CIBERONC), and 4Department of Pathology, Bellvitge University Hospital, Barcelona, Spain Date of submission 4 December 2018 Accepted for publication 19 January 2019 Published online Article Accepted 22 January 2019 Ram on y Cajal S, H€ ummer S, Peg V, Guiu X M, De Torres I, Castellvi J, Martinez-Saez E & Hernandez-Losa J (2019) Histopathology 75, 4–19. https://doi.org/10.1111/his.13828 Integrating clinical, molecular, proteomic and histopathological data within the tissue context: tissunomics Malignant tumours show a marked degree of morpho- tumour location, such as the tumour response in mela- logical, molecular and proteomic heterogeneity. This nomas versus colon carcinomas with BRAF mutations. variability is closely related to microenvironmental fac- Therefore, we understand that the correct evaluation tors and the location of the tumour. The activation of of tumours requires a system that integrates both mor- genetic alterations is very tissue-dependent and only phological, molecular and protein information in a few tumours have distinct genetic alterations. Impor- clinical and pathological context, where intratumoral tantly, the activation state of proteins and signaling heterogeneity can be assessed. Thus, we propose the factors is heterogeneous in the primary tumour and in term ‘tissunomics’, where the diagnosis will be contex- metastases and recurrences. The molecular diagnosis tualised in each tumour based on the complementation based only on genetic alterations can lead to treat- of the pathological, molecular, protein expression, ments with unpredictable responses, depending on the environmental cells and clinical data. Keywords: genomics, histopathology, proteomics, tumour heterogeneity General Background of Tumour Pathology breast cancer, expected survival is greater than 90% in patients without metastases at diagnosis and less than Despite constant progress in cancer research, mortality 25% in patients with disseminated disease.3 With this in rates remain high.1,2 Five-year survival is greater than mind, early detection, histological typing and extent of 80% for carcinoma of the breast, prostate and bladder, dissemination are critical for treatment success. Further- cutaneous melanoma and testicular and thyroid more, the more than 250 tumour types and hundreds tumours, but less than 10% in pancreatic, liver, oeso- of subtypes, on a background of thousands of genetic, phageal, lung and stomach tumours and glioblastoma. epigenetic and microRNA alterations,4–6 highlight the If we focus on tumour stage, we observe that in luminal importance of precise tumour sample evaluation in deciding the most appropriate clinical treatment. Inter- and intratumour heterogeneity (ITH)7,8 act Address for correspondence: S Ramon y Cajal, MD, PhD, as a critical barrier to improving therapeutics: most Department of Pathology, Vall d’Hebron University Hospital, of the targets and biomarkers approved by the United Autonoma University Barcelona, Passeig Vall d’Hebron 119-129, 08035, Barcelona, Spain. e-mail: sramon@vhebron.net States Food and Drug Administration are not © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Tissunomics: beyond genomic and proteomics 5 expressed uniformly throughout tumour tissue; for clones and microenvironment cells includes leucocytes, example, the human epidermal growth factor recep- lymphocytes, fibroblasts and endothelial cells. A mini- tor 2 (HER2) in gastric adenocarcinoma,9,10 proges- mum number of cells is also required for a clone to be terone and oestrogen receptors in breast tumours biologically viable – a phenomenon known as the (considered positive when detected in 1% of tumour Allee effect. This clonal cooperation has been proposed cells)11 and EML4–ALK translocation in lung adeno- in pancreatic and breast cancer and melanoma and for carcinoma (threshold of 15%),12 among others.13,14 circulating tumour cell (CTC) clusters.29–33 SOME ONCOGENIC ALTERATIONS ARE Current Situation and Challenges TISSUE-SPECIFIC WHY ARE MALIGNANT TUMOURS Certain genetic alterations are only observed in the con- HETEROGENEOUS? text of specific cell or tissue types;34 for example, The paradigm of cancer as a genetic disease has alterations in the APC gene in colorectal cancer, CDH1 existed since the early 1970s. For the last 50 years, (E-cadherin) in gastric carcinoma, the VHL gene in cancer has been explained according to the somatic renal carcinoma or BCR/ABL in leukaemia.35,36 These mutation theory (SMT),15 which states that genetic specific genetic alterations demonstrate that tissue type mutations accumulate in cells by inducing clonal plays a role, which is currently not well understood and expansion. Following Darwinian-type evolution, an ini- is under intensive investigation. The higher incidence tial clone accumulates multiple mutations, resulting in of certain somatic mutations in some cell types might malignant transformation characterised by growth be explained by a cell- or tissue-specific chromosome independence and resistance to apoptosis. These clones disposition during interphase. This, along with epige- have trunk and branch mutations, and may also have netic imprinting, may serve as a mechanism of gene driver and passenger mutations.16–18 During tumour regulation and allow differential access of carcinogens progression and after chemotherapy or radiotherapy, to certain DNA regions, resulting in a higher probability genetic alterations accumulate to select for the best- of mutations and specific rearrangements in some cells adapted clone. Molecular heterogeneity has been but not others. Similarly, the accessibility of certain described in renal adenocarcinoma, breast cancer and chromosomal regions might explain cell-type-specific glioblastoma, among others.19 It is thought that these chromosomal translocations such as BCR/ABL in leuco- clones with greater proliferative advantages and resis- cytes or VHL mutations in renal cell carcinoma.37,38 tance to local stresses then predominate.20–24 Some mutations and genetic alterations are uniformly Other theories include the tissue organisation field present in the entire organism, such as germline muta- theory (TOFT), which postulates that neoplasia arises tions, BRCA1 and BRCA2, VHL in Von Hippel Lindau from a local tissue disorder and may be due to a defect syndrome and as many as 30 others, but they are usu- in interaction between cells and other tissue compo- ally only associated with tumours in specific tissues, nents,15,25–27 where epigenetic or genetic changes such as BRCA1/2 in breast and ovarian tumours.39 affect both epithelial and stromal cells in multiple loca- This clearly indicates that driver, trigger or passenger tions. Grouping together the evidence for these theo- mutations only exert an effect in certain tissues. ries, it has been demonstrated that (i) many spontaneous somatic mutations accumulate in cells SOME HISTOPATHOLOGICAL FEATURES SUGGEST during their lifetime, (ii) oncogenes and tumour sup- ONCOGENIC ALTERATIONS pressor genes play a critical role in cancer development and (iii) environmental factors trigger cancer onset The specific morphological patterns observed in human and progression.15 Regarding field cancerisation, some tumours have helped to identify distinct genetic authors have proposed that cancer is a tissue disorder changes: for example, characteristic chromosomal with similarities to abnormal embryonic develop- translocations have been identified in desmoplastic ment,15,25 while others propose that cancer is a con- round-cell tumours, clear-cell sarcoma, synovial sar- sortium of clones and local factors.28 It is understood coma and rhabdoid tumours.40–42 Moreover, a simple that a single clone cannot harbour all the genetic alter- haematoxylin and eosin stain can indicate the presence ations required for the invasive and metastatic tumour of a number of hereditary tumours with specific genetic phenotype and that several clones acquire malignant alterations, such as medullary thyroid carcinoma, mul- properties that, synergistically, allow cells to grow, tiple endocrine neoplasia (MEN),43,44 the cribriform- invade and metastasise. This consortium of tumour cell morular variant of papillary thyroid carcinoma,45 © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 6 S Ramon y Cajal et al. familial adenomatous polyposis,46 medullary breast car- may display premalignant histological changes such cinoma, familial breast cancer associated with specific as dysplasia; furthermore, many genetic alterations BRCA1 mutations,47 renal carcinoma, associated with occur without cytological changes. These environ- SDHB mutations which shows flocculent cytoplasmic mental changes can affect tumour progression by vacuoles,48,49 hereditary leiomyomatosis and renal cell exerting evolutionary pressure on tumour cells and cancer syndrome,50 and sebaceous lesions associated ultimately determining which mutations are with Muir–Torre syndrome in colon cancer.51,52 Thus, selected.27 Examples of preneoplastic histological hereditary cancer predisposition syndromes can be iden- lesions include dysplasia (cytological atypia, large tified by immunohistochemistry, which has emerged as hyperchromatic nuclei or nucleoli) and in-situ carcino- a cost-effective strategy (see Table 1).53 mas (with no basement membrane invasion), which In contrast, other molecular alterations have been are observed in most carcinomas. It is noteworthy associated with tumours whose morphological charac- that some early neoplastic lesions such as colon ade- teristics are strikingly distinct, such as (i) the ETV6– nomas and skin naevi are polyclonal. Progression to NTRK3 translocation, detected in diverse tumour types cancer probably involves the accumulation of multiple including infantile fibrosarcoma, cellular mesoblastic field cancerisation driver mutations among synergisti- nephroma and secretory breast carcinoma;54 (ii) cally acting groups of mutations.27 In this regard, translocation affecting the ALK gene, in anaplastic molecular studies of peritumour areas in normal ducts large-cell lymphoma, lung adenocarcinoma, inflamma- of the breast or prostate may show a high number of tory myofibroblastic tumour and others;55 (iii) EWSR1– genetic alterations and driver mutations.61 The clini- CREB1 translocation in clear-cell sarcoma and cal meaning of these molecular changes remains angiomatoid fibrous histiocytoma;56 and (iv) BRAF unknown, and further large studies are needed to mutations and translocations in benign nevi, malignant establish their correlation with tumour relapse. melanoma, colon adenocarcinoma, glioblastomas and Nonetheless, two important conclusions can be pilocytic astrocytoma, as well as others.57 drawn: (i) the study of non-tumour tissue can be rele- Thus, the association between genetic alterations vant in predicting prognosis and risk of second or and specific tumour types is not clear, but some multiple neoplasia and (ii) we must be very cautious histopathological patterns suggest certain genetic making a diagnosis of malignancy based only on alterations and hereditary tumour syndromes. molecular data, as not all mutations in, for example, driver genes, are directly linked to cancer develop- ment. Problems and Barriers Mixed tumours as an example of the relevance of BIAS AND LIMITATIONS OF MOLECULAR integrating molecular features into their morphological CLASSIFICATION AND SOMATIC THEORY context Intratumour heterogeneity and pitfalls in the Mixed tumours provide the most extreme examples of interpretation of oncogenic factors in small biopsies intratumour heterogeneity. They are composed of Intratumour heterogeneity at the morphological, two or more histological components, each with a dif- molecular and proteomic level is seen in most ferent natural history, and are occasionally detected tumours. In melanoma, intratumour heterogeneity of in every organ of the human body. BRAF mutations is described in more than 10% of In most cases, the two biologically distinct tumour cases11,13,14,58 so, depending on the zone biopsied components are clonally related, and it is thought within a tumour, the determination of BRAF muta- that the more aggressive component arises from the tions may be wild-type or mutant, with clear clinical more indolent one and is usually responsible for the implications (see Figure 1). This example can be aggressiveness of the tumour. The amount of each extended to other targets and biomarkers, such as component is also closely related to outcome. EGFR mutations in non-small cell lung cancer In a small proportion of cases the two components (NSCLC),59 ALK rearrangement and others.60 are not clonally related: they are not true mixed tumours, but collision tumours in which two distinct Molecular and preneoplastic changes in the context of the neoplasms have developed independently, with a dif- adjacent tissue: field cancerisation ferent natural history. Until now, tumour diagnosis has focused on histologi- In a high proportion of mixed tumours, morpholog- cal and molecular features, yet it is well known that ical appearance correlates with molecular diversity. the surrounding cells and other areas of the tissue For example, mixed endometrioid and serous © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 7 Table 1. Molecular alterations detected by immunohistochemistry (IHC) analysis in tumour samples Molecular Principles of immunohistochemical alterations in tumors staining in tumoral cells Examples Chromosomal Overexpression of protein encoded BCL2 expression in follicular lymphoma translocation by one of the genes involved in fusion or chimeric protein encoded by the two Cyclin D1 expression in mantle cell lymphoma genes involved in fusion. Positive ALK expression in different tumour cells (anaplastic lymphoma, NSCLC) immunostaining in tumour cells NUT expression in NUT midline carcinoma ERG expression in prostate carcinoma ROS1 expression in NSCLC Gene mutation Aberrant subcellular localisation Nuclear translocation of B-catenin in (CRC, desmoid fibromatosis, of antigen or protein cribriform morular variant of PTC) Cytoplasmic expression of nucleophosmin in AML Cytoplasmic expression of BRCA1 in BC. Stabilisation and strong expression P53 in different tumour types (high grade serous ovary carcinoma) of the protein encoded by the gene mutated Mutation specific antibodies, IDH1 R132H expression in gliomas the antibody do not recognize the WT form of the protein (gene) EGFR L858R expression in NSCLC EGFR Del 19 expression in NSCLC BRAF V600E expression in different tumor types (Melanoma, PTC, CRC, others.) Presence or lack of mutation of certain LCC without IGH mutation is associate with overexpression of ZAP-70 genes give an overexpression of certain surrogate markers Gene deletion or Inactivating mutation, deletion or Loss of E-cadherin staining in lobular breast carcinoma loss of function promoter hypermethylation of gene gives a loss of protein expression Loss of INI-1 staining in rhabdoid tumor, and epithelioid sarcoma of the encoded genes Loss of staining for any MMR proteins (MLH1, MSH2, MSH6, PMS2) in HNPCC with MSI Loss of staining for parafibromin in parathyroid Carcinoma Loss of staining for Rb in spindle cell /pleomorphic lipoma Lack of staining for SDHB in hereditary paragangliomas and gastrointestinal stromal tumor Loss of PTEN expression in endometrial cancer with PTEN mutation and Cowden Syndrome Gene amplification Increase in copy number of gene HER2 strong staining in breast and gastric cancer associated with gives an overexpression of HER2 amplification the encoded protein Strong staining for MDM2 or CDK4 associates with 12q14-15 amplification in liposarcoma or well-differentiated osteosarcoma. Strong MET expression in NSCLC Adapted from Chan et al.53. © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 8 S Ramon y Cajal et al. BRAF V600E Molecular Heterogenity Pylocytic Astrocytoma Pleomorphic Xanthoastrocytoma Low Grade Gliomas Lung Adenocarcinoma Papillary Tyroid Carcinoma Gastric Carcinoma Metastatic Melanoma Nevus Erdheim Chester Disease Colorrectal Carcinoma Low Grade Ovarian Carcinoma Langerhans cell Histiocytosis Primary Melanoma Hairy Cell Leukemia Non Druggable Druggable Figure 1. Genetic alterations such as BRAF mutations can be detected in many different tumours. Importantly, the biological meaning and the response to specific BRAF inhibitors depend on tumour type and tissue localisation. carcinomas of the endometrium contain elements melanoma, the same inhibitors of the same BRAF with the microscopic characteristics of endometrioid mutations in colorectal cancer induce feedback loops and serous carcinoma, and molecular analysis of in signalling pathways that can have a detrimental microdissected tumour tissue shows that the two effect.63,64 Cell signalling pathways are also highly components have a different mutational profile. Other regulated, often dependent on numerous positive and mixed tumours are ambiguous, and different micro- negative feedback loops, which are specific to the cell scopic features develop in tumour components with type and tumour environment. For this reason, similar molecular alterations, probably as a result of BRAF inhibition in colorectal cancer indirectly interaction with the microenvironment.62 results in EGFR receptor activation and subsequent In some mixed tumours the microscopic diversity is PI3K–Akt pathway activation. Similarly, ALK alter- extreme. This is the case in biphasic synovial sar- ations and their inhibitors do not have the same coma, carcinosarcomas and some mesotheliomas. In clinical response in lung tumours as in neuroblas- carcinosarcomas (named differently in different tomas.65 Further examples are inhibitors of EGFR, organs: metaplastic carcinoma, sarcomatoid carci- HER2 and HER3, depending on the tumour type and noma, carcinosarcoma) the malignant epithelial com- location. In a recent basket clinical trial, not all ponent gives rise to mesenchymal elements, and may tumours were affected by the same HER mutations, even contain fully developed sarcomatoid features, and response to the HER2 inhibitor (neratinib) var- such as malignant skeletal muscle, cartilage or bone ied in different tissues: for example, colorectal or elements. Integration of the molecular features into bladder cancer with HER2 mutations did not the appropriate microscopic context is essential to respond.66 In summary, specific drug treatment understand these tumours and accurately assess depends on the tissue context.34 prognosis. Assessment and validation of molecular subtypes by Cancer-type-independent molecular classification of protein expression therapeutic targets Molecular classification based on mRNA expression Currently, precision oncology is based on the exis- profiling of gene groups is a frequent approach to tence of alterations in certain specific therapeutic tar- identify and classify molecular subtypes of tumours. gets regardless of tumour location or type. Tumours One of the major achievements of molecular classifi- with somatic mutations such as BRAF, EGFR or gene cation has been the subcategorisation according to rearrangements such as ALK are considered thera- prognosis, and it is instrumental for the identification peutic options regardless of tumour type or location, of new genetic alterations. However, mRNA-based and tumours may even be characterised as BRAFo- expression studies are often limited by a low repro- mas or ALKomas. However, some clinical data have ducibility, depending on mRNA stability. Further- shown that, while BRAF inhibitors are effective in more, single-cell sequencing approaches are not © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 9 feasible in routine diagnostics. Currently, samples are intermediate or low CpG island methylator phenotype sequenced as the bulk of tumour cells and therefore (CIMP)], transcriptomics and RNA array: proximal do not provide information on ITH. In addition to these proliferative, proximal inflammatory and terminal technical issues, monitoring gene expression at the respiratory unit.83 Recently, using an integrative mRNA level does not necessarily correlate with the analysis with all available omics data, six different function of a protein within a cell, as many post-tran- molecular subtypes have been described.84 Protein scriptional modifications determine whether or not markers to distinguish those subtypes are still under mRNA finally gives rise to a functional protein. Impor- way and that classification is not used in clinical tantly, however, the different molecular subtypes based oncology guides. on enormous mRNA expression matrices can be vali- Nevertheless, druggable genetic alterations can be dated according to protein or gene expression using observed in the different molecular subtypes, and cur- just a few in-situ hybridisation or IHC probes in most rent oncology guidelines follow a reasonable scheme molecular cancer subtypes (see Table 2). based on druggable genes, combining polymerase chain reaction (PCR)-based methods to identify EGFR Breast carcinoma. According to the mRNA expression and BRAF mutations, and in-situ hybridisation or profile, six molecular subtypes are described: luminal A, immunohistochemistry for the expression of BRAF luminal B, HER2+, basal-like, claudin-low and normal V600E, ALK, ROS, HER2, RET and MET. breast-like.67,68 However, RNA signatures can vary dramatically between patients with the same molecular Glioblastomas. Four molecular subtypes have been subtype69 and, with the exception of the basal subtype, described based on gene expression profile: proneural, are not highly reproducible for diagnosis on microarray mesenchymal, mixed and unknown,85 and up to six analysis.67,70,71 A high degree of ITH has been based on methylation arrays, recurrent driver genes described, as assessed by immunohistochemical (IHC) and transcriptomics, by the DKFZ (Deutsches studies,72 array-CGH studies71 and massive parallel Krebsforschungs-zentrum).86–89 These molecular sequencing technologies.73 Given these problems in the subtypes are used currently in clinical oncology guides. exact diagnosis of breast cancer subtype by molecular In brief, according to the most recent World Health characterisation, IHC-based characterisations are Organisation (WHO) classification,90 grade IV gliomas recommended as a diagnostic approach in most of the are classified as IDH-mutant GB, IDH-wt GB and consensus oncology guidelines.74 Additional IHC midline glioma with H3 K27M mutation. In order to staining helps to distinguish luminal A and B subtypes, discriminate between subtypes the following can be such as GATA3 and FOXM.75 CK5 and EGFR expression analysed using IHC, fluorescence in-situ hybridisation define the basal subtype of breast cancer,76 and CD44 (FISH) or PCR: IDH1/IDH2, ATRX, TP53, FGFR, NF1, and Snail expression determine the claudin-low H3.1 and H3.3 mutations, EGFR and PDGFR subtype.77 Finally, the normal breast-like subtype is amplifications, BRAF V600 mutation or translocations defined by low expression of the above-mentioned and CDKN2A/CDKN2B deletion. markers and expression of factors such as CD36 in a low-grade histological tumour.78 Medulloblastomas. Medulloblastomas have four molec- ular subtypes with clear clinical differences (group 1 Colorectal carcinoma. Molecular subtypes of colorectal WNT activated; group 2 SHH activated; group 3 or carcinoma have been distinguished by genomic, GABAergic; and group 4 or glutaminergic). These sub- epigenetic and transcriptomic parameters.79,80 Four types can be assessed by immunohistochemistry or FISH, different colorectal carcinomas, CMS1–4, have been studying b-catenin, GAB1, YAP1, Filamin A, p53, Glki1, described according to immunohistochemistry. The 17q+ (group 3), MYC and CDK6 amplification.91–94 molecular subtypes can be identified as mesenchymal type (CMS4), which has the worst prognosis, immune Endometrial carcinoma. Two types of endometrial type (CMS1) and epithelial subtypes (CMS2/3), based carcinoma have been described based on a combination on four proteins: CDX2, FRMD6, HTR2B and ZEB1. of epidemiological, clinical, histological and molecular In all subtypes, the drivers and actionable genes must genetic data, distinguishing type I endometrioid be studied according to approved protocols (e.g. RAS, endometrial carcinoma (EEC) and type II non- BRAF).81,82 endometrioid endometrial carcinoma (NEEC).95 A molecular classification of EC has been proposed based Lung carcinoma. Three molecular subtypes have been on The Cancer Genome Atlas Research Network described based on DNA methylation [high, (TCGA) study, published in 2013. TCGA analysis © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 10 S Ramon y Cajal et al. Table 2. Molecular signatures detected by IHC and FISH analysis in FFPE tumour samples Tumour type Molecular signature Classification Immunohistochemistry FISH Breast cancer Luminal A ER+, PR+, GATA+, FOXM+, Ki67 <15% Luminal B ER+, PR+, GATA+, FOXM+, Ki67 >15% HER2 enriched ER , PR , HER2+ ERB2 ampl. Basal-like ER , PR , HER2 , CK5+, EGFR+ Claudin low ER , PR , HER2 , CD44+, Snail+ Normal breast-like ER , PR , HER2 , CD36+ Colon cancer CMS1 (immune) FRMD6 , HTR2B , ZEB1 , CDX2+ CMS2 FRMD6 , HTR2B , ZEB1 , CDX2+ CMS3 FRMD6 , HTR2B , ZEB1 , CDX2+ CMS4 (mesenchymal) FRMD6+, HTR2B+, ZEB1+ CDX2 Glioblastomas Proneural IDH mut, p53mut, OLIG2 PDGFRA ampl, CDK4 ampl, CDK6 ampl Mesenchymal CD44, NF1 EGFR ampl Classical EGFR vIII mut, p53 EGFR ampl, loss of PTEN and CDKN2A Neural – Medulloblastoma WNT-activated b-Catenin nuclear, GAB1 , YAP1+, Filamin A+ Monosomy chr. 6 SHH-activated, TP53-WT b-Catenin cyto, GAB1+, YAP1+, Filamin A+ GLKi1 ampl, PTCH1 del SHH-activated, TP53-mut b-Catenin cyto, GAB1+, YAP1+, Filamin A+, p53+ MYCN ampl, CLI2 ampl, 17p loss Group 3 (Gabaergic) b-Catenin cyto, GAB1 , YAP1 , Filamin A 17q ampl Group 4 (Glutaminergic) b-Catenin cyto, GAB1 , YAP1 , Filamin A MYC ampl, CDK6 ampl Endometrial Ultramutated (POLE) Hypermutated (MSI) MMRd (MLH1 ) Copy number low Copy number high P53+ IHC, Immunohistochemistry; FISH, Fluorescence in-situ hybridisation; FFPE, Formalin-fixed paraffin-embedded. revealed four tumour groups:96 group 1, EEC with show different clinical, pathologic and molecular somatic inactivating mutations in POLE exonuclease features. and very high mutation rates (hypermutated) (7%); group 2, EEC with microsatellite instability (MSI), Urothelial bladder carcinoma. Different molecular frequently with MLH-1 promoter hypermethylation classifications are currently being explored. One of and high mutation rates (28%); group 3, EEC with low them, based on RNA genomewide studies, has copy number alterations (39%), also called EEC with no characterised two main molecular groups of bladder specific molecular profile; and finally, group 4 (serous- cancer with prognostic and predictive impact.97 like or copy-number high) (26%), with a low mutation Group 1 basal subtype includes a subgroup of rate but frequent TP53 mutations. These four types claudin-low and is characterised by expression of the © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 11 basal cytokeratins CK5, CK14 and CD44. Group 2 more than 8000 patients.103 The study encompassed luminal subtype has mutations of the fibroblast massive studies of sequencing, mRNA expression, pro- growth factor receptor 3 (FGFR3) gene and urothelial tein expression and engaged databases from the differentiation markers such as CK20, GATA3 and TCGA and the Human Protein Atlas. The authors uroplakins. The luminal subtype has a more also carried out in-depth bioinformatics and systems favourable prognosis than the basal subtype.97 biology analysis. Interestingly, more than 2700 genes Recently, a prospective multicentre validation study correlated with prognosis, showing opposite prognos- with a 12-gene signature progression score in non- tic effects depending on tumour type and location. muscle-invasive bladder cancer was demonstrated to Within the genes associated with worse prognosis, have independent prognostic power beyond clinical two groups were described. The first group comprised and histopathological risk factors, and could help in genes related to cell proliferation, of which more than stratifying patients to optimise treatment and follow- 190 were identified (of the 314 described related to up.98 Other molecular alterations have also been the cell cycle). This gives an idea of the great redun- described, such as FGFR3 mutations in 80%,99 dancy of gene alterations, especially the level of PI3KCA mutations in approximately 20% and expression that may exist depending on tumour type. mutations in the core promoter region of the human The second group comprised genes associated with telomerase reverse transcriptase (TERT) gene in up to tissue differentiation and their prognostic correlation. 85% of bladder tumours. Interestingly, FGFR3 and This is a point that has always been described in PIK3CA mutations have been demonstrated as pathology, distinguishing tumours with a low degree biomarkers on liquid biopsy for disease surveillance of malignancy, i.e. histologically better differentiated, for bladder cancer.100 Different cell cycle genes may from those with a high degree of malignancy, i.e. be altered in invasive urothelial carcinoma, mainly poorly differentiated with a higher degree of cellular tumour suppressor genes, including TP53, p16 and atypia. The study also found that more than 50% of Rb.101 Finally, combinations of TP53, p21, p27 and the genes associated with prognosis were not related pRb proved to be more precise in classifying bladder to Hanahan and Weinberg’s classic hallmarks.6 cancer patients into risk groups.102 Therefore, one of the conclusions was the enormous variation in gene expression profiles, both at an inter- and intratumour level. They described new sets of GENE EXPRESSION HETEROGENEITY AND genes that, according to their expression, were associ- ENVIRONMENTAL CELLS ated with prognosis in the different tumour types, mRNA expression versus protein expression underlining the clinical importance and the validity Molecular tumour subtypes based on mRNA expres- of protein expression versus mRNA or genetic alter- sion profiles often show discordance with protein ations. For example, in lung carcinoma, they pro- expression levels, depending on the RNA stability and posed a prognostic correlation with the expression of the area sampled. It is well known that RNA studies the genes ERO1A, S100A16, S100A6, MKi67, are problematic regarding reproducibility, and that SLC2A1, TACC3, ANLN, and CADM1.103–105 protein expression shows the real functional status of the targets. A classic example is breast cancer, in Assessment of intratumour heterogeneity based on protein which mRNA profiles can define at least six different expression molecular categories. While this can be used as a Currently, the most robust readout for ITH assessment general classification scheme, only IHC or FISH anal- comes from IHC. A classic example is ITH of oestrogen ysis of proteins in tumour cells within the individual receptors, which may be positive in just 1% of cells; in subgroup can provide the exact diagnosis to guide fact, this ITH has been reported to be associated with treatment and prognosis. Some reports67 describe dis- long-term risk of fatal breast cancer.106 Other exam- cord, when comparing mRNA and protein expression ples are the heterogeneous expression of cyclin D1 in levels, of up to 50% of cases in HER2 status, and in mantle cell lymphomas, which otherwise carry a more than 15% of cases when distinguishing between homogeneous CCND1–IGH translocation107 and the luminal A and luminal B. subtype of HER2+ breast or gastric tumours that can range from 10% to 100% positive cells.9,108 Tissue-specific protein expression profiles Within an entire tumour, protein expression often The relation between tissue context and protein depends on local factors such as hypoxia, nutrient expression has recently been reported in an extensive starvation or oxidative stress, independently of the gene expression study of 17 different tumour types in cell’s genetic background. These environmental cues © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 12 S Ramon y Cajal et al. can alter entire signalling pathways downstream of biopsies it is impossible to integrate the role of leuco- well-established oncogenic alterations.109 Many cytes, stromal cells and lymphocytes and the assess- tumours may have activating mutations in the most ment of immune checkpoints. upstream elements of certain signalling cascades, such as RAS, EGFR and HER2, but phosphorylation of downstream factors such as MAP-kinases or mTOR Discussion may be inhibited by environmental factors;110 there- THE SOLUTION: TO INTEGRATE CLINICAL, fore, even if specific driver genetic alterations are RADIOLOGICAL, MOLECULAR AND EXPRESSION inhibited, no effect will be observed in cells in which DATA WITHIN A TISSUE CONTEXT these signalling pathways are inhibited by such fac- tors.109,111 We postulate that cancer formation is a consortium of tumour cell clones, microenvironmental cells and local Evaluation of heterogeneity in immune checkpoints and stressors (such as hypoxia) that alter protein expres- inflammatory cells sion and enhance tumour heterogeneity. Therefore, we Primary tumours have heterogeneous epithelial cell believe that cancer must be considered within the con- content, and form an ecosystem with the environ- text of tissue type. The correct diagnosis of a tumour mental cells, including stromal cells, endothelial cells, should assimilate the clinical, radiological, histopatho- leucocytes, lymphocytes, histiocytes and their secreted logical, molecular and proteomic data. factors. In a pathological context, we can quantify As intratumour protein and molecules are highly and evaluate the expression of the immune check- heterogeneous and diagnostic biopsies are small, we points programmed cell death ligand 1 (PD-L1) and envision that correlation with radiological imaging programmed cell death-1 (PD-1) in epithelial and and nuclear medicine will be crucial to select areas inflammatory parts of the tumour and use this infor- that are more representative of the whole tumour. mation for clinical correlation studies. Similar to cell Determination of the exact histological type is cru- signalling pathways, key components of immune cial for an exact diagnosis; for example, in lung carci- checkpoints, such as PD-L1, can also have heteroge- noma there is a huge difference in prognosis and neous expression throughout the tumour.112 The PD- treatment between diagnoses of small-cell carcinoma, L1 protein is expressed in a wide range of cell types, adenocarcinoma or large-cell carcinoma. Similarly, in including lymphocytes, macrophages and dendritic ovarian carcinoma, prognosis and survival varies cells, and is stimulated by variable and complex widely between types: high- or low-grade serous car- mechanisms. In contrast to other immune checkpoint cinoma, endometrioid carcinoma, clear-cell carci- components, PD-L1 is rarely expressed in normal tis- noma or mucinous adenocarcinoma. Most tumours sue but is induced in tumour-associated tissue, is a can be diagnosed according to histological criteria as tumour biomarker and is also an attractive drug tar- high- or low-grade. The distinction is based on the get.113,114 As many current treatment schemes are degree of cell differentiation, the intensity of cellular based on immunotherapy, pathological assessment atypia and cell proliferation. A proliferation index, should include immune checkpoint proteins. Further- such as Ki67, can be used to quantify the percentage more, the heterogeneous and dynamic expression of of replicative cells within a tumour. This simple pro- factors such as PD-L1 in basal conditions and in tein can be detected regardless of which genetic response to different treatments presents a challenge alterations are inducing the proliferation. The above- to predicting which patients will respond to anti- mentioned classical factors have all been validated by PDL1 therapies. The assessment of CD4+, CD8+ cells, RNA and molecular data.103 and tumour mutational burden (TMB) can also be helpful in clinical decision-making. TMB has recently Identification of master genes and funnel factors been proposed as a new biomarker to quantify the As described above, cellular stress can alter gene potential tumoral neoantigenicity, with the aim of expression in tumour cells at several levels modifying predicting patients who will benefit from biochemical signalling pathways and feedback immunocheckpoint inhibitors. Nevertheless, different loops,111 and protein expression can be heterogenous. technologies and different thresholds have been pro- While these biochemical changes are restricted to cer- posed to define high versus low TMB in different tis- tain areas within the tumour, funnel factors such as sue samples.115–117 p4E-BP1 and peIF4E are usually expressed evenly Finally, the role of the microenvironment in car- throughout the tumour, regardless of the upstream cinogenesis is gaining importance, although in small oncogenic alterations and local environmental © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 13 changes. The search for cellular targets expressed in biomarker quantification and to develop smart and most areas of tumours is becoming crucial to over- robust algorithms. It can be based on three main come tumour heterogeneity. areas or objectives: (i) tumour type classification, (ii) Because of the huge amount of data and complex tumour prognosis (prediction system, learning from positive and negative feedback regulating the final experience) and (iii) personalised treatment allocation signal and the biological effect, systems biology is system (also a reinforcement learning system). In fact, becoming essential for the understanding and identifi- deep learning is already being applied in various cation of the real nodes or factors which are drivers tumour types, including breast, lung, melanoma and in specific tumours in individual patients. The proto- glioblastoma.123–126 The integration of such informa- oncogene myc in Burkitt lymphoma and the canoni- tion and algorithms into everyday clinical practice cal nuclear factor kappa B heterodimer in diffuse can help to minimise subjective evaluations and large B cell lymphoma are classic examples of so- improve the precise diagnosis for each tumour and called master genes.118–120 patient (see Figure 2). Systems biology encompasses tools that hold great Advances in digital pathology promise for deciphering the vulnerabilities of the Advances in digital pathology (whole slide imaging) tumour ecosystem as a whole. Studies are based on now allow interpretation of digital slides on computer the premise that multiple oncogenic events converge stations and integration of molecular features into on a number of cellular networks that may contain the microscopic context. The differential expression of essential or synthetically lethal clinically targetable genes in different areas of the tumour can be inter- hubs or factors,127,128 such as the eIF4F complex. preted precisely, and haematoxylin and eosin-stained Nonetheless, many contemporary systems biology microscopic sections can be merged with multiple approaches do not consider intratumour heterogene- immunohistochemical stains for different proteins. ity. This is relevant because key functional nodes Digital pathology also allows tumour annotation for within the metabolic, signal transduction and gene microdissection for different tumour components, expression networks responsible for supporting the with appropriate cellularity assessment of tumour ele- tumour phenotype are critically dependent on the ments and the non-neoplastic microenvironment. heterogeneity of the tumour. These data can also help Digital pathology is essential to establish comput- to classify tumours according to histopathological, erised algorithms to correlate molecular alterations biochemical and genomic features and thus help to with tumour phenotype. tailor the diagnosis and clinical management to a patient’s biological tumour profile. Accordingly, mul- Tissue-specific multiplex technology tidimensional molecular and gene expression data, Given the numerous alterations in genes, protein which are associated with the response to antitumour expression and mRNA, it is clear that we will have to treatments and clinical progress, are thought to facili- implement multiplex technologies that will allow us tate the selection of patients who are more likely to to study multiple factors simultaneously on the same respond to targeted or ‘precise’ therapies.129–131 The histological sections. With this approach we can min- availability of systemwide data in a variety of cancers imise the amount of tissue needed for molecular and is facilitating the development of approaches that go protein studies and for quantification of the expres- beyond the classic, reductionist paradigms that asso- sion of different factors and their inter-relationship. ciate single genes with cellular phenotypes and func- Different platforms are emerging, with different quan- tions. Systems biology approaches consider the titative multispectral fluorescence approaches, or with interplay between multiple molecular factors that mass spectrometry in situ, to allow the study and underpin phenotype development. Accordingly, rather quantification of several dozen markers. Also, new than a genetic disease, cancer is now perceived as a next-generation sequencing technologies are being disease of ‘networks’,131 and to fully grasp the com- implemented in routine testing combining DNA and plexity of the tumour ecosystem the networks driving RNA analysis from the same slide reducing the cancer need to be mapped and their dynamics and amount of tissue and also the ITH.121,122 evolution over time deciphered. Emerging data show that these cancer networks are constantly rewired in An interpretative and holistic data assessment with ‘deep part by clonal interactions, changing microenviron- learning’ (artificial intelligence with automatic learning) ments and the acquisition of novel molecular alter- The main objective for the use of deep learning in ations that are largely induced by anticancer pathology is to improve existing algorithms for treatments.127,132 Therefore, minor subpopulations © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 14 S Ramon y Cajal et al. Clinical features Immunostaining Asymmetry Margin S-100 Color Ki-67 Diameter p16INK4 Evolution PDL1 PET assessment Others.. Dissemination grade Age Location Histology Molecular parameters Histological type T C C A A G AT A C A Mutations Clark Fusions Breslow 20 30 Amplifications Mitosis Ulcers LOH Regression TMB Vascular invasion Others.. Satellite lesions patient data deep learning Malignant diagnosis Benign Low risk High risk (Braf WT,PDL1-5%, others...) Figure 2. Schematic representation of the potential use of artificial intelligence (AI) in the diagnosis of melanoma. All information (clinical features, histology, molecular alterations, other parameters) are analysed through AI establishing different algorithms resulting in a more precise diagnosis. that are not readily detectable in bulk tumours or Liquid biopsies that can adapt to hostile environments may emerge As biopsies are associated with a degree of invasive- following treatments that specifically target cancer- ness and cannot always be repeated, new approaches driving mutations present in the predominant tumour have been proposed, such as the analysis of circulat- subpopulations. Thus, in the coming years, systems ing tumour cells (CTCs), plasma-derived cell-free biology implementation may be crucial for (i) tumour DNA (cfDNA) or RNA and, more recently, detailed, clinically oriented subclassification of cancer DNA from circulating tumour exosomes in blood.133– 135 types, (ii) mapping of oncogenic networks and identi- This strategy enables a global assessment of the fication of the critical nodes to overcome the effects of constellation of somatic genetic alterations in a ITH and (iii) informing future preclinical research tumour irrespective of its anatomical location, and is and design of Phase I clinical trials by anticipating becoming a good alternative for the study of a therapy response in silico and predicting the best tar- tumour mutational landscape in situations where gets for each patient and tumour (personalised cancer biopsies or surgical samples are difficult to obtain. therapy or precision medicine). Hence, liquid biopsy appears to be a very relevant © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 15 clinical g protein imagin m nic sion oncoge s expres or e e ra ti o n alt th an di ag no s tic s s system gy Big data g patholo arnin digital Deep le g imagin Figure 3. Schematic drawing of a tumour section for histopathological analysis. Heterogeneity within a tumour is depicted by different col- ours of the tumoral cells. Stromal cells and infiltrating cells of the immune system are coloured in grey and orange, respectively. Variables considered for the evaluation of the tumour sample are indicated in the boxes above the slide. New methodologies, proposed to improve the diagnosis, are shown in the boxes below. tool for the identification of potential actionable geno- Conflicts of interest mic alterations and therapeutic targets, monitoring treatment response, predicting disease progression All authors declare no conflicts of interest. before clinical and radiological confirmation and iden- tifying mechanisms of resistance in the particular References context of patients with early-stage and metastatic disease.136,137 Furthermore, liquid biopsy provides 1. Global Burden of Disease Cancer Collaboration, Fitzmaurice C, additional cues on ITH, as reported in a recent pilot Allen C, Barber RM et al. Global, regional, and national can- cer incidence, mortality, years of life lost, years lived with dis- study in patients with early-stage NSCLC.138 ability, and disability-adjusted life-years for 32 cancer groups, 1990 to 2015: a systematic analysis for the Global Burden of Disease Study. JAMA Oncol. 2017; 3; 524–548. 2. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2018. CA Can- Conclusion cer J. Clin. 2018; 68; 7–30. Tumours show such a degree of morphological, 3. Yong M, Jensen AO, Jacobsen JB, Norgaard M, Fryzek JP, Sor- ensen HT. Survival in breast cancer patients with bone metas- molecular and proteomic heterogeneity, and a close tases and skeletal-related events: a population-based cohort relationship with microenvironmental factors, that study in Denmark (1999–2007). Breast Cancer Res. Treat. for the accurate pathological evaluation of tumours 2011; 129; 495–503. an integrated approach seems essential: all genomic 4. Rosai J, Ackerman LV. Rosai and Ackerman’s surgical pathology. and proteomic data should be evaluated taking into 10th ed. Maryland Heights, MI: Mosby, 2011. 5. DeVita VT Jr, Lawrence TS, Rosenberg SA et al. eds. Hellman account the heterogeneity and tissue type (Figure 3). and Rosenberg’s cancer: principles and practice of oncology. We propose the integration of all such factors, along Philadelphia: Wolters Kluwer, 2015. with the classical pathological factors, within the 6. Hanahan D, Weinberg RA. Hallmarks of cancer: the next gen- context of the tissue; we call this approach tissu- eration. Cell 2011; 144; 646–674. nomics. 7. McGranahan N, Swanton C. Biological and therapeutic impact of intratumor heterogeneity in cancer evolution. Cancer Cell 2015; 27; 15–26. 8. Rybinski B, Yun K. Addressing intra-tumoral heterogene- Acknowledgements ity and therapy resistance. Oncotarget 2016; 7; 72322– 72342. S.R.Y.C. acknowledges support from Fondo de Investi- 9. Gomez-Martin C, Concha A, Corominas JM et al. Consensus of gaciones Sanitarias (FIS; PI17/02247 and PI14/ the Spanish Society of Medical Oncology (SEOM) and Spanish 01320), Centro de Investigaci on Biomedica en Red de Society of Pathology (SEAP) for HER2 testing in gastric carci- noma. Clin. Transl. Oncol. 2011; 13; 636–651. C ancer (CIBERONC; CB16/12/00363) and Generalitat 10. Bartley AN, Washington MK, Colasacco C et al. HER2 testing de Catalunya (AGAUR; 2017 SGR 1799 and 2014 and clinical decision making in gastroesophageal adenocarci- SGR 1131). noma: guideline from the College of American Pathologists, © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 16 S Ramon y Cajal et al. American Society for Clinical Pathology, and the American 30. Cheung KJ, Ewald AJ. A collective route to metastasis: seeding Society of Clinical Oncology. J. Clin. Oncol. 2017; 35; 446–464. by tumor cell clusters. Science 2016; 352; 167–169. 11. Hammond ME, Hayes DF, Dowsett M et al. American Society 31. Cheung KJ, Padmanaban V, Silvestri V et al. Polyclonal breast of Clinical Oncology/College of American Pathologists guide- cancer metastases arise from collective dissemination of ker- line recommendations for immunohistochemical testing of atin 14-expressing tumor cell clusters. Proc. Natl Acad. Sci. estrogen and progesterone receptors in breast cancer. J. Clin. USA 2016; 113; E854–E863. Oncol. 2010; 28; 2784–2795. 32. Aceto N, Bardia A, Miyamoto DT et al. Circulating tumor cell 12. Abe H, Kawahara A, Azuma K et al. Heterogeneity of clusters are oligoclonal precursors of breast cancer metastasis. anaplastic lymphoma kinase gene rearrangement in non- Cell 2014; 158; 1110–1122. small-cell lung carcinomas: a comparative study between 33. Maddipati R, Stanger BZ. Pancreatic cancer metastases harbor small biopsy and excision samples. J. Clin. Oncol. 2015; 10; evidence of polyclonality. Cancer Discov. 2015; 5; 1086– 800–805. 1097. 13. Eckel-Passow JE, Lachance DH, Molinaro AM et al. Glioma 34. Schneider G, Schmidt-Supprian M, Rad R, Saur D. Tissue-spe- groups based on 1p/19q, IDH, and TERT promoter mutations cific tumorigenesis: context matters. Nat. Rev. Cancer 2017; in tumors. N. Engl. J. Med. 2015; 372; 2499–2508. 17; 239–253. 14. Rodon J, Saura C, Dienstmann R et al. Molecular prescreening 35. Polishchuk LA, Telegeev GD. Role of the hybrid Bcr/Abl to select patient population in early clinical trials. Nat. Rev. kinase in the pathogenesis of chronic myeloid leukemia lack- Clin. Oncol. 2012; 9; 359–366. ing C-Abl and CXCR4 proteins. Exp. Oncol. 2014; 36; 138– 15. Bertolaso M. Philosophy of cancer: a dynamic and relational view. 143. New York: Springer, 2016. 36. Marega M, Piazza RG, Pirola A et al. BCR and BCR-ABL regu- 16. Gerlinger M, Rowan AJ, Horswell S et al. Intratumor hetero- lation during myeloid differentiation in healthy donors and in geneity and branched evolution revealed by multiregion chronic phase/blast crisis CML patients. Leukemia 2010; 24; sequencing. N. Engl. J. Med. 2012; 366; 883–892. 1445–1449. 17. de Bruin EC, McGranahan N, Mitter R et al. Spatial and tem- 37. Stevens TJ, Lando D, Basu S et al. 3D structures of individual poral diversity in genomic instability processes defines lung mammalian genomes studied by single-cell Hi-C. Nature cancer evolution. Science 2014; 346; 251–256. 2017; 544; 59–64. 18. Zhang J, Fujimoto J, Zhang J et al. Intratumor heterogeneity 38. Tjong H, Li W, Kalhor R et al. Population-based 3D genome in localized lung adenocarcinomas delineated by multiregion structure analysis reveals driving forces in spatial genome sequencing. Science 2014; 346; 256–259. organization. Proc. Natl Acad. Sci. USA 2016; 113; E1663– 19. Lawrence JE, Patel AS, Rovin RA et al. Quantification of pro- E1672. toporphyrin IX accumulation in glioblastoma cells: a new 39. Nagy R, Sweet K, Eng C. Highly penetrant hereditary cancer technique. ISRN Surg. 2014; 2014; 405360. syndromes. Oncogene 2004; 23; 6445–6470. 20. Gerlinger M, Horswell S, Larkin J et al. Genomic architecture 40. Chang F. Desmoplastic small round cell tumors: cytologic, his- and evolution of clear cell renal cell carcinomas defined by tologic, and immunohistochemical features. Arch. Pathol. Lab. multiregion sequencing. Nat. Genet. 2014; 46; 225–233. Med. 2006; 130; 728–732. 21. Lloyd MC, Cunningham JJ, Bui MM, Gillies RJ, Brown JS, Gate- 41. Nielsen TO, Poulin NM, Ladanyi M. Synovial sarcoma: recent nby RA. Darwinian dynamics of intratumoral heterogeneity: discoveries as a roadmap to new avenues for therapy. Cancer not solely random mutations but also variable environmental Discov. 2015; 5; 124–134. selection forces. Cancer Res. 2016; 76; 3136–3144. 42. Guerreiro Stucklin AS, Ramaswamy V, Daniels C, Taylor MD. 22. Turajlic S, Swanton C. Metastasis as an evolutionary process. Review of molecular classification and treatment implications Science 2016; 352; 169–175. of pediatric brain tumors. Curr. Opin. Pediatr. 2018; 30; 3–9. 23. Almendro V, Kim HJ, Cheng YK et al. Genetic and phenotypic 43. Castinetti F, Moley J, Mulligan L, Waguespack SG. A compre- diversity in breast tumor metastases. Cancer Res. 2014; 74; hensive review on MEN2B. Endocr. Relat. Cancer 2018; 25; 1338–1348. T29–T39. 24. Alderton GK. Tumour evolution: epigenetic and genetic 44. Guilmette J, Nose V. Hereditary and familial thyroid tumours. heterogeneity in metastasis. Nat. Rev. Cancer 2017; 17; 141. Histopathology 2018; 72; 70–81. 25. Soto AM, Sonnenschein C. The tissue organization field theory 45. Francesco C, Paolo T, Del Vecchio MT, Monica G, Guilia M, of cancer: a testable replacement for the somatic mutation Aniello SG. Correspondence re: Cameselle-Teijeiro J, Chan theory BioEssays. News Rev. Mol. Cell. Dev. Biol. 2011; 33; JKC: cribriform-morular variant of papillary carcinoma: a dis- 332–340. tinctive variant representing the sporadic counterpart of 26. Dotto GP. Multifocal epithelial tumors and field cancerization: familial adenomatous polyposis-associated thyroid carcinoma? stroma as a primary determinant. J. Clin. Invest. 2014; 124; Mod. Pathol. 2000; 13; 363–365. 1446–1453. 46. Ma H, Brosens LAA, Offerhaus GJA, Giardiello FM, de Leng 27. Curtius K, Wright NA, Graham TA. An evolutionary perspec- WWJ, Montgomery EA. Pathology and genetics of hereditary tive on field cancerization. Nat. Rev. Cancer 2018; 18; 19–32. colorectal cancer. Pathology 2018; 50; 49–59. 28. Ramon YCS, Capdevila C, Hernandez-Losa J et al. Cancer as 47. Musolino A, Bella MA, Bortesi B et al. BRCA mutations, an ecomolecular disease and a neoplastic consortium. Bio- molecular markers, and clinical variables in early-onset chim. Biophys. Acta 2017; 186; 484–499. breast cancer: a population-based study. Breast 2007; 16; 29. Chapman A, Fernandez del Ama L, Ferguson J, Kamarashev 280–292. J, Wellbrock C, Hurlstone A. Heterogeneous tumor subpopula- 48. Ricketts C, Woodward ER, Killick P et al. Germline SDHB tions cooperate to drive invasion. Cell Rep. 2014; 8; 688– mutations and familial renal cell carcinoma. J. Natl Cancer 695. Inst. 2008; 100; 1260–1262. © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 17 49. Gill AJ, Hes O, Papathomas T, Sedivcova M et al. Succinate 66. Hyman DM, Piha-Paul SA, Won H et al. HER kinase inhibi- dehydrogenase (SDH)-deficient renal carcinoma: a morpho- tion in patients with HER2- and HER3-mutant cancers. Nat- logically distinct entity: a clinicopathologic series of 36 ure 2018; 554; 189–194. tumors from 27 patients. Am. J. Surg. Pathol. 2014; 38; 67. Eroles P, Bosch A, Perez-Fidalgo JA, Lluch A. Molecular biol- 1588–1602. ogy in breast cancer: intrinsic subtypes and signaling path- 50. Tomlinson IP, Alam NA, Rowan AJ et al. Germline mutations ways. Cancer Treat. Rev. 2012; 38; 698–707. in FH predispose to dominantly inherited uterine fibroids, skin 68. Heng YJ, Lester SC, Tse GM et al. The molecular basis of leiomyomata and papillary renal cell cancer. Nat. Genet. breast cancer pathological phenotypes. J. Pathol. 2017; 241; 2002; 30; 406–410. 375–391. 51. Winship IM, Dudding TE. Lessons from the skin – cutaneous 69. Califano A, Alvarez MJ. The recurrent architecture of tumour features of familial cancer. Lancet Oncol. 2008; 9; 462–472. initiation, progression and drug sensitivity. Nat. Rev. Cancer 52. Ponti G, Losi L, Di Gregorio C et al. Identification of Muir- 2017; 17; 116–130. Torre syndrome among patients with sebaceous tumors and 70. Lusa L, McShane LM, Reid JF et al. Challenges in projecting keratoacanthomas: role of clinical features, microsatellite clustering results across gene expression-profiling datasets. J. instability, and immunohistochemistry. Cancer 2005; 103; Natl Cancer Inst. 2007; 99; 1715–1723. 1018–1025. 71. Geyer FC, Weigelt B, Natrajan R et al. Molecular analysis 53. Chan JK, Ip YT, Cheuk W. The utility of immunohistochem- reveals a genetic basis for the phenotypic diversity of meta- istry for providing genetic information on tumors. Int. J. Surg. plastic breast carcinomas. J. Pathol. 2010; 220; 562–573. Pathol. 2013; 21; 455–475. 72. Potts SJ, Krueger JS, Landis ND et al. Evaluating tumor 54. Argani P, Fritsch M, Kadkol SS, Schuster A, Beckwith JB, heterogeneity in immunohistochemistry-stained breast cancer Perlman EJ. Detection of the ETV6–NTRK3 chimeric RNA of tissue. Lab. Invest. 2012; 92; 1342–1357. infantile fibrosarcoma/cellular congenital mesoblastic 73. Ding L, Wendl MC, McMichael JF, Raphael BJ. Expanding the nephroma in paraffin-embedded tissue: application to chal- computational toolbox for mining cancer genomes. Nat. Rev. lenging pediatric renal stromal tumors. Mod. Pathol. 2000; Genet. 2014; 15; 556–570. 13; 29–36. 74. Goldhirsch A, Winer EP, Coates AS et al. Personalizing the 55. Schram AM, Chang MT, Jonsson P, Drilon A. Fusions in solid treatment of women with early breast cancer: highlights of tumours: diagnostic strategies, targeted therapy, and acquired the St Gallen International Expert Consensus on the Primary resistance. Nat. Rev. Clin. Oncol. 2017; 14; 735–748. Therapy of Early Breast Cancer 2013. Ann. Oncol. 2013; 24; 56. Cantile M, Marra L, Franco R et al. Molecular detection and 2206–2223. targeting of EWSR1 fusion transcripts in soft tissue tumors. 75. Tang P, Tse GM. Immunohistochemical surrogates for molec- Med. Oncol. 2013; 30; 412. ular classification of breast carcinoma: a 2015 update. Arch. 57. Hall RD, Kudchadkar RR. BRAF mutations: signaling, epi- Pathol. Lab. Med. 2016; 140; 806–814. demiology, and clinical experience in multiple malignancies. 76. Denkert C, Liedtke C, Tutt A, von Minckwitz G. Molecular Cancer Control 2014; 21; 221–230. alterations in triple-negative breast cancer - the road to new 58. Valachis A, Ullenhag GJ. Discrepancy in BRAF status among treatment strategies. Lancet 2017; 389; 2430–2442. patients with metastatic malignant melanoma: a meta-analy- 77. de Beca FF, Caetano P, Gerhard R et al. Cancer stem cells sis. Eur. J. Cancer 2017; 81; 106–115. markers CD44, CD24 and ALDH1 in breast cancer special his- 59. Chen ZY, Zhong WZ, Zhang XC et al. EGFR mutation hetero- tological types. J. Clin. Pathol. 2013; 66; 187–191. geneity and the mixed response to EGFR tyrosine kinase inhi- 78. Hagemann IS. Molecular testing in breast cancer: a guide to bitors of lung adenocarcinomas. Oncologist 2012; 17; 978– current practices. Arch. Pathol. Lab. Med. 2016; 140; 815– 985. 824. 60. Cai W, Lin D, Wu C et al. Intratumoral heterogeneity of ALK- 79. Guinney J, Dienstmann R, Wang X et al. The consensus rearranged and ALK/EGFR coaltered lung adenocarcinoma. J. molecular subtypes of colorectal cancer. Nat. Med. 2015; 21; Clin. Oncol. 2015; 33; 3701–3709. 1350–1356. 61. Myers MB, Banda M, McKim KL, Wang Y, Powell MJ, Parsons 80. Dienstmann R, Vermeulen L, Guinney J, Kopetz S, Tejpar S, BL. Breast cancer heterogeneity examined by high-sensitivity Tabernero J. Consensus molecular subtypes and the evolution quantification of PIK3CA, KRAS, HRAS, and BRAF mutations of precision medicine in colorectal cancer. Nat. Rev. Cancer in normal breast and ductal carcinomas. Neoplasia 2016; 18; 2017; 17; 79–92. 253–263. 81. Marisa L, de Reynies A, Duval A et al. Gene expression classi- 62. Coenegrachts L, Garcia-Dios DA, Depreeuw J et al. Mutation fication of colon cancer into molecular subtypes: characteriza- profile and clinical outcome of mixed endometrioid-serous tion, validation, and prognostic value. PLoS Med. 2013; 10; endometrial carcinomas are different from that of pure 1001453. endometrioid or serous carcinomas. Virchows Arch. 2015; 82. Trinh A, Trumpi K, De Sousa EMF et al. Practical and 466; 415–422. robust identification of molecular subtypes in colorectal can- 63. Prahallad A, Sun C, Huang S et al. Unresponsiveness of colon cer by immunohistochemistry. Clin. Cancer Res. 2017; 23; cancer to BRAF(V600E) inhibition through feedback activa- 387–398. tion of EGFR. Nature 2012; 483; 100–103. 83. Ringner M, Jonsson G, Staaf J. Prognostic and chemother- 64. Karoulia Z, Gavathiotis E, Poulikakos PI. New perspectives for apy predictive value of gene-expression phenotypes in pri- targeting RAF kinase in human cancer. Nat. Rev. Cancer mary lung adenocarcinoma. Clin. Cancer Res. 2016; 22; 2017; 17; 676–691. 218–229. 65. Claeys S, Denecker G, Cannoodt R et al. Early and late effects 84. Chen F, Zhang Y, Parra E et al. Multiplatform-based molecu- of pharmacological ALK inhibition on the neuroblastoma lar subtypes of non-small-cell lung cancer. Oncogene 2017; transcriptome. Oncotarget 2017; 8; 106820–106832. 36; 1384–1393. © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. 18 S Ramon y Cajal et al. 85. Verhaak RG, Hoadley KA, Purdom E et al. Integrated genomic bladder: a contemporary series from the Bladder Cancer analysis identifies clinically relevant subtypes of glioblastoma Research Consortium. J. Urol. 2006; 176; 2414–2422; dis- characterized by abnormalities in PDGFRA, IDH1, EGFR, and cussion 22. NF1. Cancer Cell 2010; 17; 98–110. 103. Uhlen M, Zhang C, Lee S et al. A pathology atlas of the 86. Aldape K, Zadeh G, Mansouri S, Reifenberger G, von Deimling human cancer transcriptome. Science 2017; 357; pii: A. Glioblastoma: pathology, molecular mechanisms and eaan2507. markers. Acta Neuropathol. 2015; 129; 829–848. 104. Thomas A, Liu SV, Subramaniam DS, Giaccone G. Refining 87. Korshunov A, Schrimpf D, Ryzhova M et al. H3-/IDH-wild the treatment of NSCLC according to histological and molecu- type pediatric glioblastoma is comprised of molecularly and lar subtypes. Nat. Rev. Clin. Oncol. 2015; 12; 511–526. prognostically distinct subtypes with associated oncogenic dri- 105. Rotow J, Bivona TG. Understanding and targeting resistance vers. Acta Neuropathol. 2017; 134; 507–516. mechanisms in NSCLC. Nat. Rev. Cancer 2017; 17; 637–658. 88. Reifenberger G, Wirsching HG, Knobbe-Thomsen CB, Weller 106. Lindstrom LS, Yau C, Czene K et al. Intratumor heterogeneity M. Advances in the molecular genetics of gliomas – implica- of the estrogen receptor and the long-term risk of fatal breast tions for classification and therapy. Nat. Rev. Clin. Oncol. cancer. J. Natl Cancer Inst. 2018; 110; 726–733. 2017; 14; 434–452. 107. Jares P, Colomer D, Campo E. Molecular pathogenesis of man- 89. Sturm D, Witt H, Hovestadt V et al. Hotspot mutations in tle cell lymphoma. J. Clin. Invest. 2012; 122; 3416–3423. H3F3A and IDH1 define distinct epigenetic and biological sub- 108. Apicella M, Corso S, Giordano S. Targeted therapies for gas- groups of glioblastoma. Cancer Cell 2012; 22; 425–437. tric cancer: failures and hopes from clinical trials. Oncotarget 90. Louis DN, Perry A, Reifenberger G et al. The 2016 World 2017; 8; 57654–57669. Health Organization classification of tumors of the central 109. Ramon YCS, Castellvi J, Hummer S, Peg V, Pelletier J, Sonen- nervous system: a summary. Acta Neuropathol. 2016; 131; berg N. Beyond molecular tumor heterogeneity: protein syn- 803–820. thesis takes control. Oncogene 2018; 37; 2490–2501. 91. Ramaswamy V, Remke M, Bouffet E et al. Risk stratification of 110. Ramon YCS, De Mattos-Arruda L, Sonenberg N, Cortes J, Peg childhood medulloblastoma in the molecular era: the current V. The intra-tumor heterogeneity of cell signaling factors in consensus. Acta Neuropathol. 2016; 131; 821–831. breast cancer: p4E-BP1 and peIF4E are diffusely expressed 92. Ellison DW, Dalton J, Kocak M et al. Medulloblastoma: clinico- and are real potential targets. Clin. Transl. Oncol. 2014; 16; pathological correlates of SHH, WNT, and non-SHH/WNT 937–941. molecular subgroups. Acta Neuropathol. 2011; 121; 381–396. 111. Martinez A, Sese M, Losa JH et al. Phosphorylation of eIF4E 93. Kaur K, Kakkar A, Kumar A et al. Integrating molecular sub- confers resistance to cellular stress and DNA-damaging classification of medulloblastomas into routine clinical prac- agents through an interaction with 4E-T: a rationale for tice: a simplified approach. Brain Pathol. 2016; 26; 334–343. novel therapeutic approaches. PLoS One 2015; 10; 94. Clifford SC, Lusher ME, Lindsey JC et al. Wnt/Wingless path- e0123352. way activation and chromosome 6 loss characterize a distinct 112. Mahoney KM, Rennert PD, Freeman GJ. Combination cancer molecular sub-group of medulloblastomas associated with a immunotherapy and new immunomodulatory targets. Nat. favorable prognosis. Cell Cycle 2006; 5; 2666–2670. Rev. Drug Discov. 2015; 14; 561–584. 95. Bokhman JV. Two pathogenetic types of endometrial carci- 113. Chen DS, Mellman I. Oncology meets immunology: the can- noma. Gynecol. Oncol. 1983; 15; 10–17. cer-immunity cycle. Immunity 2013; 39; 1–10. 96. Cancer Genome Atlas Research Network, Kandoth C, Schultz 114. Madore J, Vilain RE, Menzies AM et al. PD-L1 expression in N, Cherniack AD et al. Integrated genomic characterization of melanoma shows marked heterogeneity within and between endometrial carcinoma. Nature 2013; 497; 67–73. patients: implications for anti-PD-1/PD-L1 clinical trials. Pig- 97. Seiler R, Ashab HAD, Erho N et al. Impact of molecular sub- ment Cell Melanoma Res. 2015; 28; 245–253. types in muscle-invasive bladder cancer on predicting 115. Hellmann MD, Ciuleanu TE, Pluzanski A et al. Nivolumab response and survival after neoadjuvant chemotherapy. Eur. plus ipilimumab in lung cancer with a high tumor muta- Urol. 2017; 72; 544–554. tional burden. N. Engl. J. Med. 2018; 378; 2093–2104. 98. Dyrskjot L, Reinert T, Algaba F et al. Prognostic impact of a 116. Chalmers ZR, Connelly CF, Fabrizio D et al. Analysis of 12-gene progression score in non-muscle-invasive bladder 100,000 human cancer genomes reveals the landscape of cancer: a prospective multicentre validation study. Eur. Urol. tumor mutational burden. Genome Med. 2017; 9; 34. 2017; 72; 461–469. 117. Yarchoan M, Hopkins A, Jaffee EM. Tumor mutational bur- 99. Tomlinson DC, Baldo O, Harnden P, Knowles MA. FGFR3 pro- den and response rate to PD-1 inhibition. N. Engl. J. Med. tein expression and its relationship to mutation status and 2017; 377; 2500–2501. prognostic variables in bladder cancer. J. Pathol. 2007; 213; 118. Faumont N, Durand-Panteix S, Schlee M et al. c-Myc and 91–98. Rel/NF-kappaB are the two master transcriptional systems 100. Christensen E, Birkenkamp-Demtroder K, Nordentoft I et al. activated in the latency III program of Epstein-Barr virus- Liquid biopsy analysis of FGFR3 and PIK3CA hotspot muta- immortalized B cells. J. Virol. 2009; 83; 5014–5027. tions for disease surveillance in bladder cancer. Eur. Urol. 119. Schmitz R, Ceribelli M, Pittaluga S, Wright G, Staudt LM. 2017; 71; 961–969. Oncogenic mechanisms in Burkitt lymphoma. Cold Spring 101. van Rhijn BW, van der Kwast TH, Vis AN et al. FGFR3 and Harb. Perspect. Med. 2014; 4; pii: a014. P53 characterize alternative genetic pathways in the patho- 120. Guo X, Koff JL, Moffitt AB et al. Molecular impact of selective genesis of urothelial cell carcinoma. Cancer Res. 2004; 64; NFKB1 and NFKB2 signaling on DLBCL phenotype. Oncogene 1911–1914. 2017; 29; 4224–4232. 102. Shariat SF, Karakiewicz PI, Palapattu GS et al. Outcomes of 121. Letovanec I, Finn S, Zygoura P et al. Evaluation of NGS and radical cystectomy for transitional cell carcinoma of the RT-PCR methods for ALK rearrangement in European NSCLC © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19. Tissunomics: beyond genomic and proteomics 19 patients: results from the European Thoracic Oncology Plat- 130. Gatenby RA, Cunningham JJ, Brown JS. Evolutionary triage form Lungscape Project. J. Thorac. Oncol. 2018; 13; 413–425. governs fitness in driver and passenger mutations and sug- 122. Oliveira DM, Mirante T, Mignogna C et al. Simultaneous iden- gests targeting never mutations. Nat. Commun. 2014; 5; tification of clinically relevant single nucleotide variants, copy 5499. number alterations and gene fusions in solid tumors by tar- 131. Krogan NJ, Lippman S, Agard DA, Ashworth A, Ideker T. geted next-generation sequencing. Oncotarget 2018; 9; The cancer cell map initiative: defining the hallmark net- 22749–22768. works of cancer. Mol. Cell 2015; 58; 690–698. 123. Han Z, Wei B, Zheng Y, Yin Y, Li K, Li S. Breast cancer mul- 132. Paul JM, Templeton SD, Baharani A, Freywald A, Vizea- ti-classification from histopathological images with structured coumar FJ. Building high-resolution synthetic lethal net- deep learning model. Sci. Rep. 2017; 7; 4172. works: a ‘Google map’ of the cancer cell. Trends Mol. Med. 124. Vandenberghe ME, Scott ML, Scorer PW, Soderberg M, Bal- 2014; 20; 704–715. cerzak D, Barker C. Relevance of deep learning to facilitate 133. Alix-Panabieres C, Pantel K. Clinical applications of circulat- the diagnosis of HER2 status in breast cancer. Sci. Rep. 2017; ing tumor cells and circulating tumor DNA as liquid biopsy. 7; 45938. Cancer Discov. 2016; 6; 479–491. 125. Yu KH, Zhang C, Berry GJ et al. Predicting non-small cell 134. Buder A, Tomuta C, Filipits M. The potential of liquid biop- lung cancer prognosis by fully automated microscopic pathol- sies. Curr. Opin. Oncol. 2016; 28; 130–134. ogy image features. Nat. Commun. 2016; 7; 12474. 135. De Mattos-Arruda L, Cortes J, Santarpia L et al. Circulating 126. Rathore S, Akbari H, Rozycki M et al. Radiomic MRI signa- tumour cells and cell-free DNA as tools for managing breast ture reveals three distinct subtypes of glioblastoma with dif- cancer. Nat. Rev. Clin. Oncol. 2013; 10; 377–389. ferent clinical and molecular characteristics, offering 136. Murtaza M, Dawson SJ, Tsui DW et al. Non-invasive analysis prognostic value beyond IDH1. Sci. Rep. 2018; 8; 5087. of acquired resistance to cancer therapy by sequencing of 127. Werner HM, Mills GB, Ram PT. Cancer systems biology: a plasma DNA. Nature 2013; 497; 108–112. peek into the future of patient care? Nat. Rev. Clin. Oncol. 137. Murtaza M, Dawson SJ, Pogrebniak K et al. Multifocal clonal 2014; 11; 167–176. evolution characterized using circulating tumour DNA in a 128. Kiel C, Serrano L. Structural data in synthetic biology case of metastatic breast cancer. Nat. Commun. 2015; 6; approaches for studying general design principles of cellular 8760. signaling networks. Structure 2012; 20; 1806–1813. 138. Jamal-Hanjani M, Wilson GA, Horswell S et al. Detection of 129. Russo M, Siravegna G, Blaszkowsky LS et al. Tumor hetero- ubiquitous and heterogeneous mutations in cell-free DNA geneity and lesion-specific response to targeted therapy in col- from patients with early-stage non-small-cell lung cancer. orectal cancer. Cancer Discov. 2016; 6; 147–153. Ann. Oncol. 2016; 27; 862–867. © 2019 Authors. Histopathology Published by John Wiley & Sons Ltd, Histopathology, 75, 4–19.