EGFR gene amplification is relatively common and associates with outcome in intestinal adenocarcinoma of the stomach, gastro-oesophageal junction and distal oesophagus

The study population in this retrospective study consists of 220 patients diagnosed with intestinal adenocarcinoma of the stomach, gastro-oesophageal junction or distal oesophagus at the Turku University Hospital between the years 1993 and 2012. Initially, we used the clinical database of Auria Biobank (see below) to find all patients with the diagnosis of adenocarcinoma of the stomach, gastro-oesophageal junction or distal oesophagus (n = 437). The original histopathological information regarding these samples was then obtained to compile a preliminary list of patients, and the respective histological slides were retrieved from the archive. The exclusion criteria for this study were: diffuse or neuroendocrine histological subtype (n = 155), metastatic adenocarcinoma from a different organ (n = 6), intramucosal carcinoma (Tis) (n = 23) and insufficient sample material (n = 33). All cases were reanalysed by an expert gastrointestinal pathologist and the intestinal histological subtype of the tumours was confirmed by the presence of well-defined glandular structures in accordance with the Laurén classification [1]. Primarily, tissue samples from primary surgical specimens were included. In order to attain a comprehensive study population, representative biopsies were used in case of 22 patients (10 %): four (1.8 %) patients were not operated due to stage IV disease at the time of diagnosis and 18 (8.2 %) patients had received perioperative chemoradiotherapy resulting in insufficient surgical material for immunohistochemical analysis. The type of surgery was total gastrectomy for 120 (54.5 %) patients, subtotal gastrectomy or tumour resection for 79 (35.9 %) patients and palliative surgery for 17 (7.7 %) patients. The residual tumour classification was determined as R0 (no residual tumour) for 167 (75.9 %) patients, R1 (microscopic residual tumour) for 24 (10.9 %) patients and R2 (macroscopic residual) for 17 (7.7 %) patients. The residual tumour status could not be determined for 12 (5.5 %) patients. The median follow-up time for all patients was 10.5 years. The patient characteristics are presented in Table 1.

Tumour stage was assessed according to the current WHO Classification manual [9]. The study was conducted in accordance with the Declaration of Helsinki and the Finnish legislation for the use of archived tissue specimens and associated clinical information. The clinical data were retrieved, and the histological samples were collected and analysed with the endorsement of the National Authority for Medico-Legal Affairs and The Ethics Committee of the Hospital District of Southwest Finland as well as with the permission of Auria Biobank hosting the specimen archive. All the specimens were from Auria biobank, which has obtained its archived diagnostic sample collection with an opt-out procedure according to the Finnish biobank act [10]. Biobanks authorized and inspected by National Supervisory Authority for Welfare and Health can provide human specimens collected during diagnostic procedures and associated clinical information for research purposes based on the biobank’s scientific board review. Thus, informed consent from surviving patients was not required.

For each tumour, the most representative formalin-fixed paraffin-embedded (FFPE) tissue block was chosen and new sections were cut for both IHC staining and SISH. The methods for EGFR IHC and EGFR SISH have been described previously [7], and HER2 IHC was performed similarly with monoclonal HER2 antibody (clone 4B5, Ventana Medical Systems/Roche Diagnostics, Tucson, AZ, USA). HER2/Chr17 double-SISH was detected with HER2 DNA Probe and INFORM Chromosome 17 Probe (Ventana/Roche) and performed with ultraView SISH Detection Kit and ultraView Alkaline Phosphatase (AP) Red ISH Detection Kit (Ventana/Roche).

With EGFR, tumour scoring was based on the most intense membranous or membranous + cytoplasmic staining (0, negative; 1+, weak; 2+, moderate; 3+, strong). Strong staining was seen as intense reaction with 5x objective magnification, moderate staining was clearly identified with 5x objective magnification and weak staining was identified only with 10x objective magnification. Specimens were classified as IHC high if showing 2+ or 3+ membranous or membranous + cytoplasmic staining intensity in ≥10 % of tumour cells in surgical specimens or in ≥5 clustered tumour cells in biopsies. These IHC high samples were further analysed with SISH. This algorithm is based on our previous observation that high EGFR IHC staining intensity positively correlates with increased EGFR GCN [7]. With HER2 IHC, tumours were scored according to standard criteria [11, 12] and specimens showing 2+ or 3+ membranous staining in ≥10 % of tumour cells or in ≥5 clustered tumour cells in biopsies were classified as IHC high and analysed with SISH. EGFR and HER2 IHC and GCN were scored independently by two observers (EB and JS) without knowledge of the clinical information. Consensus scoring was used in case of differing individual results.

EGFR was quantified from the areas of high EGFR IHC intensity as described previously [7, 8]. Forty tumour cells with the highest number of copies were analysed from the EGFR SISH slides and an average value was calculated for each surgical sample. If these forty cells contained numerous overlapping EGFR SISH signals (clusters), the tumour was determined to have EGFR gene amplification. In biopsies, a group of ≥5 tumours cells with gene clusters was considered as amplification. One EGFR cluster was approximated to contain ≥10 gene copies. HER2 GCN was detected with chromosome 17 (Chr-17) number (number of copies of chromosome per cell) and the HER2/Chr-17 ratio was assessed according to standard criteria [13]. If HER2 gene clusters were detected in ≥10 % of tumour cells in surgical specimens or in a group of ≥5 tumour cells in biopsies, the tumour was determined to contain HER2 gene amplification. One HER2 cluster was counted as ≥6 gene copies. To validate our method of including only tumours with high EGFR IHC intensity for EGFR SISH, we assessed EGFR GCN in fifteen randomly selected tumours in which EGFR IHC was scored as negative/weak. No EGFR amplification was found in these tumours (GCN 2.1–3.3).

Statistical analyses were performed with IBM SPSS Statistics for Windows, version 21.0 (IBM Corporation, Armonk, NY). Frequency table data were analysed using the χ ² test, either with the Pearson χ ² test or Fisher’s exact test for categorical variables. 2 × 2 tables were used to calculate odds ratios (OR). Kaplan-Meier method and log-rank test as well as Cox’s proportional hazards regression model were used for univariate survival analysis. Multivariate survival analysis was performed by Cox’s proportional hazards regression model. Variables with a p-value under 0.2 in univariate analysis were included in the multivariate analyses. Time to recurrence (TTR) was calculated from the time of diagnosis to the time of first recurrence, death of primary cancer or to the last follow-up date. Only recurrences occurring ≥6 months after diagnosis were considered relevant. Earlier detection of a local or distant recurrence was considered likely to present an initially advanced disease. Patients treated with surgery or surgery and adjuvant therapy without disease recurrence ≥6 months after diagnosis were considered curatively treated. Cancer-specific survival (CSS) was calculated from the time of diagnosis to the time of death of primary cancer or the last follow-up date and overall survival (OS) from the time of diagnosis to the time of death of any cause or the last follow-up date. Five patients (2.3 %) who had received trastuzumab treatment for recurrent cancer were excluded from the CSS and OS analyses and additionally 14 patients with stage IV disease (6.4 %) from the TTR analysis. All statistical tests were two-sided and p-values under 0.05 were considered statistically significant.

All 220 tumour samples were analysed with EGFR and HER2 IHC. High membranous or membranous + cytoplasmic EGFR IHC staining intensity (2+/3+) was observed in 72 (32.7 %) of the tumours, while 2+/3+ HER2 IHC staining intensity was present in 31 (14.1 %) tumours. Among these, concurrent high IHC staining intensity of EGFR and HER2 was detected in 14 (6.4 %) tumours. The results from EGFR and HER2 IHC stainings are shown in Table 2.

Gene copy numbers were analysed with EGFR or HER2 SISH in all tumours with high EGFR or HER2 IHC staining intensity. EGFR gene amplification was found in 31/72 tumours (14.1 % of the whole study material) and HER2 gene amplification in 29/31 tumours (13.2 % of the whole study material). Among these, EGFR and HER2 co-amplification was detected in 8/14 tumours (3.6 % of the whole study material). EGFR and HER2 gene amplification status was significantly concordant in antrum (Fisher’s exact test, p = 0.004). The results from EGFR and HER2 SISH stainings according to anatomical location are presented in Table 3. There was marked intratumoural heterogeneity of EGFR and HER2 gene amplification, as shown in Figs. 1 and 2.

Evaluated by IHC staining intensity, moderate or strong EGFR protein expression was associated with the depth of tumour invasion (pT3–pT4 versus pT1–pT2; Fisher’s exact test, p = 0.029); OR 2.15, 95 % CI: 1.11–4.17), but did not associate with tumour location (distal oesophagus/GOJ/cardia versus gastric corpus/antrum/pylorus; Fisher’s exact test, p = 0.054). In contrast, no significant association was found between HER2 protein expression levels and the depth of tumour invasion or tumour location. No significant association was observed between EGFR or HER2 protein expression levels and patient gender, tumour stage or histological differentiation grade.

EGFR gene amplification was associated with deep invasion (pT3–pT4 versus pT1–pT2; Fisher’s exact test, p = 0.020; OR 3.49, 95 % CI: 1.17–10.4) and it was more commonly detected in stage III–IV tumours than in stage I–II tumours (Fisher’s exact test, p = 0.024; OR 2.55, 95 % CI: 1.18–5.51). Additionally, EGFR gene amplification was more common in tumours of distal oesophagus (5/20 tumours, 25.0 %) and GOJ/cardia (13/63 tumours, 20.6 %) than in those of gastric corpus (2/65 tumours, 3.1 %) (χ ², p = 0.013). This distribution pattern was also seen in male patients (χ ², p = 0.034) but not in female patients. When tumour location was considered as a dichotomous variable, EGFR gene amplification was still more common in proximally located tumours (distal oesophagus/GOJ/cardia versus gastric corpus/antrum/pylorus; (Fisher’s exact test, p = 0.016); OR 2.64, 95 % CI: 1.22–5.73). When analysed separately for males and females, the association between EGFR gene amplification and proximal tumours was significant in males (Fisher’s exact test, p = 0.011; OR 3.58, 95 % CI: 1.37–9.36) but not in females. In contrast, HER2 gene amplification status was not significantly associated with the depth of tumour invasion, tumour stage or tumour location. No significant association was found between EGFR or HER2 gene amplification status and patient gender, age at diagnosis or histological differentiation grade of the tumour. The association between EGFR and HER2 protein expression as well as gene amplification and different clinicopathological variables are presented in Table 4.

In univariate survival analysis, EGFR gene amplification was associated with shortened time to recurrence (TTR, median) (22 vs. 57 months, log-rank test, p = 0.026; Cox test, p = 0.028, HR: 1.73, 95 % CI: 1.06–2.83) and with shortened cancer-specific survival (CSS, median) (29 vs. 57 months, log-rank test, p = 0.033; Cox test, p = 0.035, HR: 1.67, 95 % CI: 1.04–2.69) (Fig. 3). Median TTR and CSS of the patients were both 45 months. HER2 gene amplification was not significantly associated with TTR, but patients with HER2 gene amplification had a notably lower median CSS of 22 months than patients without HER2 amplification (46 months). However, the difference was not statistically significant (log-rank test, p = 0.256) (Fig. 3).

In univariate analysis, increasing depth of tumour invasion was associated with decreased TTR and CSS (TTR: log-rank test, p < 0.0001; Cox test, p < 0.0001, HR 1.46, 95 % CI: 1.19–1.80 and CSS: log-rank test, p < 0.0001; Cox test, p < 0.0001, HR 1.60, 95 % CI: 1.30–1.96). Similarly, increasing tumour stage was associated with decreased TTR and CSS (TTR: log-rank test, p = 0.005; Cox test, p = 0.001, HR 1.52, 95 % CI: 1.18–1.96 and CSS: log-rank test, p < 0.0001; Cox test p < 0.0001, HR 1.94, 95 % CI: 1.53–2.45). In addition, increasing patient age at the time of diagnosis was associated with shorter CSS (Cox test, p = 0.048, HR 1.02, 95 % CI: 1.00 − 1.04), but not with TTR (Cox test, p = 0.341). No significant association was found between patient gender (log-rank test, TTR: p = 0.372; CSS: p = 0.818) or tumour location (log-rank test, TTR: p = 0.057; CSS: p = 0.262). In Kaplan-Meier analysis, histological differentiation grade was not associated with survival (grade I versus II versus III; log-rank test, TTR: p = 0.118; CSS: p = 0.053). However, when analysed separately grade II tumours were associated with shorter TTR in comparison to grade I tumours (univariate Cox test, p = 0.043, HR 1.95, 95 % CI: 1.02–3.74). Additionally, grade II and III tumours were associated with shorter CSS in comparison to grade I tumours (univariate Cox test, grade II: p = 0.020, HR 2.22, 95 % CI: 1.13–4.36; grade III: p = 0.029, HR 2.15, 95 % CI: 1.08–4.27). No significant association was observed between EGFR or HER2 gene amplification status and overall survival (OS). EGFR or HER2 protein expression, evaluated by IHC staining intensity, was not significantly associated with TTR, CSS or OS.

In the multivariate model for TTR, EGFR gene amplification was analysed together with tumour stage, histological differentiation grade and tumour location. In the multivariate analysis for CSS, EGFR gene amplification was analysed together with tumour stage, histological differentiation grade and patient age at the time of diagnosis. Tumour stage remained as a single predictive factor for TTR (Cox test, stage III: p = 0.014, HR 2.05, 95 % CI: 1.16–3.63) as well as for CSS (Cox test, stage III: p = 0.023, HR 1.99, 95 % CI: 1.10–3.61; stage IV: p < 0.0001, HR 11.4, 95 % CI: 5.34–24.4). The results from univariate and multivariate survival analyses are presented in Table 5.