Gastric cancer: immunohistochemical classification of molecular subtypes and their association with clinicopathological characteristics

Gastric cancer is traditionally divided into intestinal and diffuse histological subtypes, but recent molecular analyses have led to novel classification proposals based on genomic alterations. While the intestinal- and diffuse-type tumours are distinguishable from each other at the molecular level, intestinal-type tumours have more diverse molecular profile. The technology required for comprehensive molecular analysis is expensive and not applicable for routine clinical diagnostics. In this study, we have used immunohistochemistry and in situ hybridisation in molecular classification of gastric adenocarcinomas with an emphasis on the intestinal subtype. A tissue microarray consisting of 244 gastric adenocarcinomas was constructed, and the tumours were divided into four subgroups based on the presence of Epstein-Barr virus, TP53 aberrations and microsatellite instability. The intestinal- and diffuse-type tumours were separately examined. The distribution of EGFR and HER2 gene amplifications was studied in the intestinal-type tumours. Epstein-Barr virus positive intestinal-type tumours were more common in male patients (p = 0.035) and most often found in the gastric corpus (p = 0.011). The majority of the intestinal-type tumours with TP53 aberrations were proximally located (p = 0.010). All tumours with microsatellite instability showed intestinal-type histology (p = 0.017) and were associated with increased overall survival both in the univariate (p = 0.040) and multivariate analysis (p = 0.015). In conclusion, this study shows that gastric adenocarcinomas can be classified into biologically and clinically different subgroups by using a simple method also applicable for clinical diagnostics.

four subgroups based on the presence of Epstein-Barr virus, TP53 aberrations and microsatellite instability. The intestinaland diffuse-type tumours were separately examined. The distribution of EGFR and HER2 gene amplifications was studied in the intestinal-type tumours. Epstein-Barr virus positive intestinal-type tumours were more common in male patients (p = 0.035) and most often found in the gastric corpus (p = 0.011). The majority of the intestinal-type tumours with TP53 aberrations were proximally located (p = 0.010). All tumours with microsatellite instability showed intestinal-type histology (p = 0.017) and were associated with increased overall survival both in the univariate (p = 0.040) and multivariate analysis (p = 0.015). In conclusion, this study shows that Virchows Archiv conforms to the ICMJE recommendation for qualification of authorship. The ICMJE recommends that authorship be based on the following four criteria: • Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; AND • Drafting the work or revising it critically for important intellectual content; AND • Final approval of the version to be published; AND • Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Eva-Maria Birkman and Naziha Mansuri have equal contribution.

Introduction
Gastric cancer is one of the major causes of cancer-related death worldwide [1]. Gastric adenocarcinomas have traditionally been divided into intestinal and diffuse subtypes according to the Laurén classification based on the histological characteristics of the tumours [2]. However, The Cancer Genome Atlas Consortium (TCGA) has recently proposed a molecular subtyping of gastric adenocarcinomas based on the presence of Epstein-Barr virus (EBV), microsatellite instability (MSI), genomic stability (GS) and chromosomal instability (CIN) [3]. The majority of the GS tumours have diffuse histology, while the other subgroups contain predominantly intestinal-type tumours. An alternative classification has been proposed by the Asian Cancer Research Group (ACRG). This proposal stratifies gastric adenocarcinomas into tumours with MSI, microsatellite-stable tumours showing epithelial to mesenchymal transition (MSS/EMT), MSS tumours with intact TP53 activity (MSS/TP53+) and MSS tumours with functional loss of TP53 (MSS/TP53−) [4]. These classification systems have provided valuable information about the variability in biological characteristics among gastric adenocarcinomas. Instead of considering gastric cancer as a single disease, it has also become clear that when exploring new cancer therapies, future studies need to be conducted among defined sets of patients whose tumours have specific genomic abnormalities. In clinical practice, HER2 is the only predictive biomarker for targeted therapy currently used for patient selection in gastric cancer.
The complex methodologies used in these abovementioned studies are not applicable for routine clinical diagnostics, and some more straightforward methods have already been proposed [5][6][7][8]. In this study, we have constructed a nextgeneration tissue microarray (ngTMA) from 244 intestinaland diffuse-type adenocarcinomas of the stomach, gastrooesophageal junction (GOJ) and distal oesophagus [9]. Using this cohort, we have been able to identify four subgroups of tumours with distinct molecular and clinicopathological characteristics by combining the Laurén classification, MSI and TP53 immunohistochemistry (IHC) and EBER in situ hybridisation (ISH). Additionally, the analysis of Ecadherin expression was performed in order to compare our method with the previously published studies, and the distribution of EGFR and HER2 gene amplifications were examined among these subgroups in the intestinal-type tumours.

Patients and tumour specimens
The study population consists of 244 patients diagnosed with adenocarcinoma of the stomach, GOJ or distal oesophagus at the Turku University Hospital between years 1993 and 2012. The initial search in the clinical database of Auria Biobank retrieved tumour specimens from 437 patients. The exclusion criteria were carcinoma in situ (Tis, n = 23), insufficient sample material or indeterminate histology (n = 63), neuroendocrine histology (n = 3) and metastatic adenocarcinoma from a different organ (n = 6). This resulted in 190 patients with intestinal-type tumours and 152 patients with diffuse-type tumours. From these, all intestinal-type tumours and 54 representative diffuse-type tumours were included in the ngTMA. Among all of the patients, 11.9% (29/244) did receive preoperative chemotherapy (13 patients with intestinal-type and 16 with diffuse-type tumours). The type of surgery was total gastrectomy for 154 (63.1%) patients, subtotal gastrectomy or tumour resection for 72 (29.5%) patients and palliative surgery for 18 (7.4%) patients. The extent of surgery was determined as R0 (no residual tumour) for 180 (73.8%) patients, R1 (microscopic residual tumour) for 34 (13.9%) patients and R2 (macroscopic residual tumour) for 20 (8.2%) patients. The extent of surgery could not be determined for 10 (4.1%) patients. Tumour stage was assessed according to the current WHO Classification manual [10]. The median follow-up time of the patients was 125 months. All corresponding H&E slides have been reviewed for confirmation of diagnosis and adequacy of material. Relevant clinical information has been gathered from each case. The reporting of the study has been performed in compliance with the current recommendations [11]. The patient characteristics are presented in Table 1.

Tissue microarray construction
The ngTMA was created as follows [9]. The appropriate formalin-fixed paraffin-embedded (FFPE) tissue specimens were chosen based on clinical data and retrieved from the pathology archives. A representative haematoxylin-eosin (H&E) section containing areas of invasive carcinoma was selected from each tumour. New H&E slides were produced, scanned (Pannoramic P250, 3DHistech) and uploaded into the university digital microscopy web portal (casecenter.utu.fi). Each slide was viewed using Pannoramic Viewer software (3DHistech). Using the 1.0-mm annotation tool, annotations of different colours corresponding to various histological areas were placed onto each digital slide. Two annotations were placed in the centre of the tumour and two annotations in the periphery or invasive front of the tumour. The corresponding tissue cores were then transferred into the TMA blocks using an automated TMA instrument (TMA Grandmaster, 3DHistech) by overlaying each annotated digital slide with the corresponding tissue specimen. One tissue core containing benign tissue was selected from each tumour to act as a control. The constructed TMA blocks were sectioned, stained, scanned and uploaded into the web portal (casecenter.utu.fi), and each individual spot was scored by two pathologists (EB and NM). The resulting scores were combined with the clinical data for statistical analysis.
The link https://seafile.utu.fi/d/7c4aa1964b/ contains examples of our TMA results. The directory BTMA staining^contains low resolution images of all stainings performed on TMA block number 7, and BTMA map and score^contains the map and scores of the same TMA block. The directory BExample scoring^contains selected high resolution images of positive/negative and aberrant/wildtype stained tissue cores.

Immunohistochemistry and in situ hybridisation
IHC reactions were performed on 4-μm paraffin sections of each tumour on the TMA slides with BenchMark XT (Ventana/Roche). For TP53, a ready-to-use antibody clone Bp53-11 (Ventana/Roche) was used, and the protocol included mild (30 min) CC1 pretreatment together with 28-min antibody incubation. Signal detection was performed with ultraView universal DAB Detection Kit (Ventana/Roche). For MLH1, an antibody clone G168-15 (BD Pharmingen) was used at 1:5 dilution together with standard (60 min) CC1 pretreatment and 36-min antibody incubation. The signal was detected with ultraView Universal DAB Detection Kit and amplification kit. For MSH2, an antibody clone G219-1129 (BD Pharmingen) was used at 1:200 dilution together with standard CC1 pretreatment and 28-min antibody incubation. The signal was detected with ultraView Universal DAB Detection Kit. For MSH6, an antibody clone EP49 (Epitomoc) was used at 1:200 dilution together with standard CC1 pretreatment and 32-min antibody incubation. The signal was detected with ultraView Universal DAB Detection Kit. For PMS2, a ready-to-use antibody clone EPR3947 (Ventana/ Roche) was used together with extended (90 min) CC1 pretreatment, 44-min antibody incubation and 12-min HQ LINKER and HRP MULTIMER enhancements. The signal was detected with OptiView Universal DAB Detection Kit and amplification kit. For EBV, a ready-to-use EBER (EBV- GOJ gastro-oesophageal junction a Diffuse-type tumours are included in the tumours of the corpus encoded small RNA) probe (Ventana/Roche) was used together with ISH-Protease 3 pretreatment for 28 min and 1-h probe incubation. The signal was detected with ISH iVIEW Blue Detection Kit. For E-cadherin, an antibody clone NCH-38 (Agilent Technologies) was used at 1:100 dilution together with standard CC1 pretreatment and 32-min antibody incubation. The signal was detected with ultraView Universal DAB Detection Kit and amplification kit. A complete loss of or strong diffuse TP53 nuclear positivity was classified as aberrant TP53 expression. EBER ISH was scored either positive or negative according to the nuclear reaction. A tumour was classified as MSI if at least one of the markers (MLH1, MSH2, MSH6 and PMS2) showed a complete loss of nuclear reactivity together with positive background reaction in benign epithelium, smooth muscle cells and lymphocytes. Negative nuclear reactivity with negative background was considered controversial and not used for classification Loss of membranous reactivity or only faint cytoplasmic reaction was classified as aberrant E-cadherin expression (Fig. 1a). The methods for EGFR and HER2 IHC and silver in situ hybridisation (SISH) have been described previously [12][13][14][15]. In short, with EGFR the scoring was based on the most intense membranous or membranous+cytoplasmic staining (0, negative; 1+, weak; 2+, moderate; 3+, strong  Fig. 1b).
EGFR and HER2 in the intestinal-type adenocarcinomas EGFR and HER2 protein expression levels were evaluated in 183 intestinal-type adenocarcinomas. Moderate/ strong EGFR protein expression was found in 59 (32.2%) of the tumours and 25 (13.7%) of the tumours had moderate/strong HER2 protein expression. Among these, EGFR gene amplification was detected in 27/59 (14.8% of the whole study material) tumours and HER2 gene amplification in 24/25 (13.1% of the whole study material) tumours. No significant associations were observed between EGFR/HER2 protein expression level or gene amplification and TP53/EBV/MSI status. The majority of the EGFR (n = 17) or HER2 (n = 15) gene amplifications were found in the group of EBV negative and MSS tumours with aberrant TP53. Most of the co-amplifications (n = 5) were also found in this subgroup (Fig. 1b). The co-localisation of aberrant TP53 expression and either EGFR or HER2 gene amplification was detected more often in the proximal (distal oesophagus/GOJ/cardia) than distal (corpus/antrum/pylorus) intestinal-type tumours (Fisher's exact test, p = 0.019; OR 2.83, 95% CI 1.21-6.61).     (Table 4).

Discussion
In this study, we describe a straightforward method for molecular classification of gastric cancer applicable for both clinical diagnostics and research purposes. With an emphasis on the intestinal-type adenocarcinomas, we have used IHC and ISH to define four subgroups of gastric adenocarcinomas with distinct molecular and clinical characteristics.
The recent molecular profiling studies have mainly been conducted in either Western [5] or Asian populations [4,[6][7][8]. The TCGA study contains patients from various geographical regions and among them they did not find any significant difference in the prevalence of the different molecular subtypes between the East-Asian group (Vietnam and South Korea) and the overall group [3]. Among the intestinal-type tumours, the frequency of EBV positive adenocarcinomas in our Finnish study population that represents Caucasian genetic background was 9.1%, which is in line with the results of 3.0-9.5% from other studies showing no clear difference between the geographical regions [3,4,6,8]. The frequency of intestinal-type MSI tumours was 10.2% in our cohort which is similar to 9.2% found by Kim et al. (2016) but somewhat less than 24.5-26.0% found in other studies with mixed patient populations [3,4]. The presence of aberrant TP53 expression was found in 55.4% of the intestinal-type tumours in our study. This is close to the 53.1% of the TCGA study but somewhat different from the 31.3% found by Cristescu et al. (2015) or 67.3% by Kim et al. (2016). Still, no clear difference can be seen between the different geographical regions.
The molecular analyses of gastric adenocarcinomas have shown that the intestinal-and diffuse-type tumours are distinguishable from each other also at the molecular level, and the intestinal-type tumours have more diverse molecular profiles than the diffuse-type tumours, which are the predominant subtype in the Bgenomically stable^category [3,4]. Therefore, the intestinal-type adenocarcinomas have been the main focus of our analyses, and a subset of diffuse-type tumours has served as a reference group for other publications. We have shown that the intestinal-type tumours rarely contain aberrant E-cadherin expression; and that by beginning with the Laurén classification, we could concentrate 25/28 (89.3%) of the tumours with aberrant E-cadherin expression into the diffuse subgroup. Notably, none of the EBV negative, MSS and TP53 wt intestinal-type tumours showed aberrant E-cadherin expression. This suggests that some other biomarker could be more specific than E-cadherin in detecting the intestinal-type tumours not characterised by EBV positivity, MSI or TP53 aberrations. In all, these observations imply that the   histological subtype can be used as a starting point for a clinically applicable method for molecular classification. Among the diffuse-type tumours, the comparison of our results to the other studies should be considered as suggestive as we have analysed only a subset of the diffuse-type tumours diagnosed during the study period. The frequency of EBV positivity has been reported to be 3.8-9.8% among diffusetype tumours [3,4,6,8], but in our study none of the diffusetype tumours was found to contain EBV positivity. Similarly, we did not detect MSI in the diffuse-type tumours. In other studies, the proportion of MSI diffuse-type tumours has been reported to be relatively low (3.8-8.7%) [3,6] with the exception of 16.9% by Cristescu et al. (2015). The frequency of aberrant TP53 expression in our study was 19.2% among the diffuse-type tumours, which is quite similar to the observed frequencies of 26.1-27.5% [3,4] but notably different from 53.8% reported by Kim et al. (2016).
We could demonstrate the association of EBV positivity with male patients and the location of the tumour in the gastric corpus as previously shown by Ahn et al. (2017). Additionally, EBV positivity was found to be associated with poor histological differentiation among the intestinal-type tumours. Distally located tumours were more often characterised by MSI than proximal tumours, which confirms the results obtained by the TCGA and Cristescu et al. (2015). We could also show that EGFR and HER2 gene amplifications were most common among the intestinal-type tumours with EBV negativity, MSS and TP53 aberration, of which 17.3% (17/98) were EGFR and 15.3% (15/98) HER2 amplified. Notably, also among the EBV negative, MSS and TP53 wild-type tumours, the proportion of these amplifications, 15.4% (8/52) for both genes, was substantial. In the TCGA study population, 10.6% (9/85) of the EBV negative, MSS and TP53 aberrant tumours contained EGFR and 34.1% (29/ 85) contained HER2 gene amplification. Similar to our results, among the EBV negative, MSS and TP53 wild-type tumours, the proportion of EGFR gene amplification was 11.1% (5/45) and HER2 gene amplification 13.3% (6/45) [3,16,17]. In our study, the co-localisation of aberrant TP53 expression together with EGFR or HER2 gene amplification was noticed to be more common in the proximally than distally located intestinal-type tumours, which is in line with the results from a recent characterisation of oesophageal carcinomas showing strong genomic similarities between oesophageal adenocarcinomas and CIN-type gastric adenocarcinomas according to the TCGA classification [18]. Additionally, we observed that patients with MSI tumours had longer survival than patients with MSS tumours both in the univariate and multivariate analysis, which is consistent with earlier findings [4,7].
Our classification method concentrated on the intestinal subtype of gastric adenocarcinoma, where receptor tyrosine kinase (RTK) copy number alterations are more common than in diffuse-type tumours. In the TCGA study population, all tumours with either EGFR or HER2 gene amplification were observed to have intestinal-type histology [3]. However, we also examined a subset of diffuse-type tumours in order to make comparisons with the results obtained by other studies. Most likely, due to this selection, we did not find EBV positivity or MSI among the diffuse-type tumours in our study population.
In order to increase the reliability of the TMA analysis, we have examined four tissue cores from each tumour. For the EGFR and HER2 SISH, we have used whole slide sections in order to account for the potential spatial heterogeneity of the gene amplifications. Additionally, we have examined the expression of all four MSI markers instead of only one in order to increase the reliability of the classification into MSI or MSS tumours. Especially in the MSI classification, there is variation in the reported prevalence of MSI phenotype, which may partly be related to technical issues in performing and interpreting the IHC reactions. Moreover, some variability is inevitably related to using TP53 IHC as a surrogate marker for TP53 gene mutations.
Our classification method differs from that suggested by Park et al. (2016) in that our initial division was based on the Laurén classification. After that, we continued by categorising the tumours first by EBV and second by MSI reactivity. Finally, we considered the presence of TP53 aberrations.
Our method was able to determine four distinct subgroups among the intestinal-type tumours without considerable overlapping of the molecular markers. Among the diffuse-type tumours in our material, EBV positivity, MSI and aberrant TP53 expression were mutually exclusive. Among the intestinal tumours, only one EBV positive tumour showed aberrant TP53 expression and four tumours with MSI showed aberrant TP53 expression. EBV positivity and MSI were mutually exclusive also among the intestinal-type tumours, which is in line with other studies [3,6]. Only 3/183 intestinal-type tumours (1.6%) had aberrant E-cadherin expression, and even they could be classified as either EBV positive, MSI or TP53 aberrant. The proportion is comparable to the observed 4.1% (8/196) of the intestinal-type tumours with E-cadherin mutations in the TCGA study.
A few articles have recently been published describing different algorithms implementing the molecular classification system proposed by TCGA into clinical practice [5][6][7][8].
However, only few studies have utilised the Laurén classification and none of them have based their proposed algorithm on the histological subtype of the tumours. In one study, differentially expressed genes were compared by microarray analysis between intestinal and diffuse gastric cancers, and a 40-gene signature was created to serve as a prognostic tool [19]. As RTK amplifications are known to be more prevalent among the intestinal-type tumours, the histological subtype could be a relevant factor to take in to account when investigating new RTK-targeting therapies for gastric cancer [3]. In only one of these recent studies, the evaluation of both EGFR and HER2 gene amplification status has been performed by ISH [7]. However, the SISH procedure was carried out on TMA slides, which potentially does not account for the tumour heterogeneity. So far, anti-EGFR antibodies have not shown survival benefit in phase III trials including gastric or oesophagogastric cancer patients [20,21]. One reason for this may be that the screening methods used have not been able to identify the subgroup of patients possibly responsive to anti-EGFR treatment. In future clinical trials, the application of new classification methods combining both histological and molecular information may become necessary in order to improve the clinical benefit obtained by new targeted therapies.
In conclusion, we have demonstrated that gastric adenocarcinomas can be classified into biologically and clinically relevant subgroups by using a straightforward and clinically applicable method based on the Laurén classification together with immunohistochemistry and in situ hybridisation.