Gastrointestinal tissue‐based molecular biomarkers: a practical categorisation based on the 2019 World Health Organization classification of epithelial digestive tumours

Molecular biomarkers have come to constitute one of the cornerstones of oncological pathology. The method of classification not only directly affects the manner in which patients are diagnosed and treated, but also guides the development of drugs and of artificial intelligence tools. The aim of this article is to organise and update gastrointestinal molecular biomarkers in order to produce an easy‐to‐use guide for routine diagnostics. For this purpose, we have extracted and reorganised the molecular information on epithelial neoplasms included in the 2019 World Health Organization classification of tumours. Digestive system tumours, 5th edn.


Introduction
In the era of molecular medicine, with the expansion of digital pathology and the revolution in artificial intelligence (AI), molecular biomarker classifications of cancer are more important than ever before. 1 A rational cancer taxonomy is necessary to standardise diagnoses, make decisions on biomarker/drug development, and generate an appropriate background for AI tools. 2 Molecular biomarkers have a prominent role in oncological pathology, with diagnostic, predictive and/or prognostic value. A single and unified classification for biomarkers is important in order to collect relevant information and keep it updated. This represents a challenge for modern (morphomolecular) pathologists. 3 The value of biomarkers in routine tissue diagnostics A 'biomarker' is defined as 'any substance, structure, or process that can be measured in the body or its products and influence or predict the incidence of outcome or disease'. 4 Every year, there are between 15 000 and 20 000 new scientific articles on cancer biomarkers. 5 Unfortunately, of every 100 such biomarkers, <1% make it into a form that is useful for patient diagnosis or stratification, 6 mostly for a variety of scientific and technical reasons. 7 As a result, there is no clear-cut evidence in the literature regarding which biomarkers are essential for diagnostics and/or therapeutic decision-making. However, the World Health Organization (WHO) classification of tumours consensus of international experts represents the best indication of how relevant these biomarkers are in routine diagnostic practice. Our goal is to summarise the use of these biomarkers in the gastrointestinal system (from the oesophagus to the anal canal), and obtain indications of the specific weight, form and relevance of biomarker analysis in disease taxonomy and clinical decision-making.

Current biomarker classification and proposed subcategorisation
A biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal/ pathogenic processes or pharmacological therapeutic responses. 8 The 2019 World Health Organization classification of tumours. Digestive system tumours, 5th edn. includes diagnostic, predictive and prognostic molecular biomarkers as the major categories. 9 Diagnostic biomarkers are intended to help pathologists establish a specific diagnosis; predictive markers indicate the probability of patients benefiting from a specific therapy; and prognostic markers determine the outcomes of patients, in the absence of specific treatments. 8 The decision on which group a biomarker belongs to represents the first step of assessment. Some biomarkers may fulfil the criteria for more than one category. Additionally, it is possible to subdivide categories, improving their organisation and comprehension. In this manner, for a diagnostic biomarker, the second step is to determine whether it is useful in differential diagnosis or whether it contributes to cancer classification. For predictive biomarkers, the following question should be asked. Are there definitive randomised clinical trials or cohort studies that support their efficacy? If the answer is yes, then these would correspond to established predictive biomarkers. If the answer is negative, but they are currently under investigation, one may classify them as potentially predictive biomarkers. If they are not yet associated with any clinical trial, we propose to label them with the term 'preclinical predictive biomarkers'. In this circumstance, it is unlikely that they would be placed within one of the first-level groups within the WHO classification of tumours, although they often have relevance to the understanding of tumour pathogenesis, and may be included under this topic. For prognostic biomarkers, the main question is whether they are specific prognostic markers for a certain entity, or whether they are used to create risk stratification groups. Biomarkers that do not fit into any of these categories should be classified as 'others'. This classification is summarised in Figure 1.
However, in order to understand, adequately categorise and subcategorise these biomarkers, it is necessary to methodically evaluate other attributes associated with them.
Variables to consider in the categorisation of biomarkers Context is a relevant aspect to consider in any biomarker assessment. A specific biomarker can have different attributes according to the location (system/ organ) and the disease (entity) in question. For example, the presence of epidermal growth factor receptor (EGFR) gene activating mutations in oesophageal squamous cell carcinoma (SCC) is an adverse prognostic factor, and EGFR-targeted therapies have failed to improve survival. [10][11][12] The same molecular alterations in non-small-cell lung carcinoma (NSCLC) confer a better prognosis and also provide the patient with an opportunity to receive tyrosine kinase inhibitor therapy with a significant chance of improved survival. 13 Activating mutations of v-raf murine sarcoma viral oncogene homologue B1 (BRAF) in colorectal carcinoma (CRC) may confer resistance to anti-EGFR therapy and help to establish a worse prognosis, 14 whereas the same alterations (BRAF V600E activating mutation) in melanoma predict response to treatment with BRAF inhibitors, such as vemurafenib. 15 Sometimes, the existence of certain molecular alterations is known, but there are no clinical trials available to support their routine use. On other occasions, support exists for a specific organ, but not for others. An example of this is human epidermal growth factor receptor 2 gene (HER2; ERBB2) mutation in small-intestine adenocarcinoma, which can be detected. In theory, a patient with cancer harbouring the mutation will obtain benefit from anti-ERBB2 therapy, but it does not yet have an established predictive value. 16 On the other hand, alterations in the same biomarker in oesophagus/ oesophagogastric junction adenocarcinoma are predictive of response to this targeted therapy. 17

S T A T U S : S P E C I F I C M O L E C U L A R A L T E R A T I O N S
Evaluation of the status of a biomarker implies specification of the molecular alterations that give it clinical utility. Determination of the specific alteration (activating mutation, translocation, overexpression, etc.) conceptually corresponds to the exact molecular phenomenon involved. In this manner, the lack of KRAS/NRAS activating mutations in CRC predicts a favourable response to anti-EGFR therapy. 14 In contrast, the presence (but not the lack) of activating KIT mutations in other malignant neoplasms, such as melanoma or gastrointestinal stromal tumours, predicts response to imatinib therapy. 18

Gastrointestinal system biomarkers update G E N E R A L
A total of 54 different biomarkers are mentioned 98 times across the gastrointestinal tract chapters of the WHO blue book. Figure 2 summarises them, showing whether the technology used corresponds to immunohistochemical (IHC) tests, transcriptomic tools, or DNA-based mutational assays. Microsatellite instability (MSI) is described 11 times: one for diagnostic, four for prognostic and six for therapeutic purposes. p53/TP53 is mentioned 10 times: three times as an IHC test, and seven as a DNA-based mutational analysis tool. KRAS study is indicated on six occasions, all of which correspond to DNA mutational analysis with predictive utility, with one use as a potential prognostic biomarker. In contrast, overexpression of ERBB2 is indicated five times, all involving detection by the use of IHC analysis, but with the consideration that some cases will require confirmation with fluorescence insitu hybridisation. The organ with the most biomarkers mentioned was the small intestine/ampulla, with 31 different markers being mentioned, the vast majority being specific prognostic markers with potential for future use. Finally, the large intestine has more established biomarkers for current routine use than any other anatomical site, with 15 markers of diagnostic, established predictive and prognostic use.
The biomarkers discussed below are summarised by category, subcategory and organ in Tables 1 and 2.

O E S O P H A G U S
The routinely used molecular biomarkers in oesophageal lesions include the presence of aberrant IHC expression of p53, which may be associated with better diagnostic reproducibility for dysplasia (differential diagnosis biomarker) and an increased risk of neoplastic progression (prognostic risk stratification biomarker), in the context of Barrett's oesophagus. 26,27 In addition, ERBB2 overexpression and/or ERBB2 amplification in lower oesophagus/oesophagogastric junction adenocarcinoma carries a predictive value for response to ERBB2-targeted therapy (established predictive biomarker). 17 There are other markers that are not yet used in routine pathological analysis, but may be important in the near future. EGFR overexpression in oesophageal SCC is considered to be an adverse specific prognostic factor, 10 because targeted therapies have not been successful in improving survival. 11,12 The loss of mismatch repair protein expression (with the consequent MSI) and overexpression of programmed deathligand 1 (PD-L1) in oesophagus/oesophagogastric junction adenocarcinoma are linked to the potential use of immune checkpoint inhibitors. These checkpoint inhibitors are undergoing clinical trial evaluation for immunotherapy, 28 as is cytotoxic Tlymphocyte antigen 4 overexpression (potentially predictive biomarkers). 28,29 Other possible biomarkers are methylation of the cyclin-dependent kinase inhibitor 2A gene (CDKN2A) (p16) promoter, which inhibits its gene expression, and genomic instability (specifically copy number alterations). Both biomarkers have potential value as prognostic risk stratification biomarkers in Barrett dysplasia, 27 but this is not yet supported by strong retrospective or prospective studies.

S T O M A C H
The molecular biomarkers with current use in gastric tumours include ERBB2 overexpression and/or ERBB2 amplification in gastric adenocarcinoma and gastric undifferentiated carcinoma, with established predictive value for response to ERBB2-targeted therapy, 17,30,31 and the presence of the MALAT1-GLI1 fusion gene for diagnostic confirmation of gastroblastoma, a rare gastric biphasic tumour that has recently been described (differential diagnosis biomarker). 32 TP53 and retinoblastoma gene 1 (RB1) mutations also act as differential diagnosis biomarkers, helping to distinguish gastric neuroendocrine carcinomas (NECs) from G3 neuroendocrine tumours (NETs), in which these genes are more frequently wild type; this is also applicable for the remainder of the digestive organs. 33 A plethora of markers are used in gastric cancer biology, specifically with regard to specific prognostic markers, with little direct routine application. These include EGFR and c-MET overexpression, 30 MSI, 34 Epstein-Barr virus detection, 35 high expression levels of epidermal growth factor/transforming growth factor-a, vascular endothelial growth factor-A and CD44, reduced expression of E-cadherin, expression of matrix metalloproteinase (MMP)-1, MMP7, and MMP10, upregulation of SPC18 (SEC11A), and protocadherin B9 (PCDHB9) overexpression. [36][37][38] MSI and PD-L1 expression are potential predictive molecular biomarkers under investigation in clinical trials. 34,39,40 Finally, for gastric dysplasia, there are biomarkers of disease progression that are seldom used routinely today. These are: DNA content abnormalities (aneuploidy or elevated 4N fraction) 41 ; aberrant p53/TP53; mutations of RNF43, the adenomatous polyposis coli gene (APC), the AT-rich interactive domain-containing protein 1A gene (ARID1A), and the AT-rich interactive domain-containing protein 2 gene (ARID2) 42,43 ; and inactivation by promoter methylation of p16 and MLH1 (with consequent MSI). 44,45

S M A L L I N T E S T I N E A N D A M P U L L A
In ampullary and non-ampullary adenocarcinomas, only MSI is a regularly used molecular biomarker. Its indications include immunotherapy selection (established predictive biomarker), determination of a possible hereditary origin (differential diagnosis biomarker), and as a specific prognostic parameter (early results show that MSI may improve overall survival). 46,47 Markers with potential specific prognostic value, but without regular pathological use, include KRAS activating mutations in ampullary adenocarcinoma, 19 chromosome 18 deletion, chromosome 14 gain, whole arm copy number variations, and cyclin-dependent kinase inhibitor 1B gene (CDKN1B) mutations in small intestine and ampullary neuroendocrine neoplasms. [48][49][50] Other biomarkers with potential specific prognostic or predictive preclinical value in non-ampullary adenocarcinoma are TP53, the isocitrate dehydrogenase gene (IDH), the cadherin-1 gene (CDH1), the fibroblast growth factor receptor 2 gene (FGFR2), the fmslike tyrosine kinase 3 gene (FLT3), the nucleophosmin gene (NPM1), the phosphatase and tensin homologue gene (PTEN), c-MET, AKT1, RET, the Notch homologue 1 translocation-associated gene (NOTCH1), the human epidermal growth factor receptor 4 gene (ERBB4), the chimerin 2 gene (CHN2), KRAS, the SMAD family member 4 gene (SMAD4), ERBB2, and CTNNB1/E-cadherin. 16,[51][52][53][54] A P P E N D I X Multiple studies have been conducted in appendiceal adenocarcinoma, but there is insufficient evidence to make firm recommendations regarding potentially predictive and preclinical biomarkers [KRAS, MSI, and the guanine nucleotide-binding protein G gene (GNAS)]. [55][56][57] As in the rest of the digestive system, the presence of TP53 and RB1 mutations can help to    The molecular markers routinely used in CRC comprise the lack of activating mutations of KRAS/NRAS and BRAF (extended RAS testing), both of which have established predictive value for effective response to anti-EGFR therapy. 14,58 BRAF activating mutations have differential diagnostic utility in the exclusion of Lynch syndrome, and they are associated with an adverse specific prognosis. 59,60 MSI is an established predictive marker in colorectal adenocarcinoma: it is associated with a significant response to PD-L1 inhibitors for patients in whom conventional therapy has failed, confers a good prognosis to BRAF wild-type patients (specific prognostic marker), and is useful in Lynch syndrome diagnosis. 61 In addition, two different methods are being used for colorectal adenocarcinoma molecular diagnostic classification: genomic-scale analysis (hypermutated or non-hypermutated colorectal cancers) 62 and transcriptomic profiling (consensus molecular subtypes for colorectal cancer). 21 Both have potential for subtypebased targeted therapies.
4,5-bisphosphate 3-kinase catalytic subunit-a gene (PIK3CA) activating mutations are associated with a worse specific prognosis, a negative predicted response to anti-EGFR therapy, and a successful adjuvant response to acetylsalicylic acid. 63,64 Immune-related markers such as the Immunoscore represent a potential prognostic stratification tool, 65,66 and gene expression signatures have more restricted prognostic use, specifically for determining the risk of recurrence after surgery. 67,68 A N A L C A N A L The molecular biomarkers with clinical utility are p16 expression and polymerase chain reaction determination of high-risk human papilloma virus (HPV) genotypes (usually 16) in anal squamous dysplasia. The risk of progression from a high-risk lesion to SCC is influenced by HPV genotype, immune status, and other factors. [69][70][71] Other markers with potential and/or preclinical predictive utility include PD-L1 expression in SCC 72,73 and KRAS and NRAS and MSI in anal adenocarcinoma. 74

Conclusion
Despite the significant knowledge of the molecular basis of cancers of the digestive tract, there are relatively few biomarkers with established clinical utility, and most target common tumour types. Our review follows the new WHO approach to molecular markers, which is easily identifiable and can also be readily revisited when new information becomes available in the future. This systematic approach to the characterisation of new molecular markers may be used for the future taxonomy of cancers, which are also likely to benefit from computational pathology, especially within the next 5-year cycle of the WHO classification of tumours.