Esophagectomies were performed for patients with EC at the Department of Surgery, University of Cologne, Germany. Among 146 patients with locally advanced EC (cT3-4, cNx and M₀), tissue samples collected at the Institute for Pathology, University Hospital of Cologne were applied in our restrospective study. All specimens were used in accordance with the local policies of the Institutional Review Board of the University Hospital of Cologne. TNM staging was performed according to the criteria of the International Union Against Cancer (UICC) (17). Clinical staging consisted of endoscopy, endoscopic ultrasound, barium swallow, CT scanning of the abdomen and thorax and positron-emission tomography. Morphologic assessment of tumor regression was performed by an objective histopathologic examination as described previously (16). The resected specimens were fixed in formalin (10%). The whole tumor region including a safety margin was excised by a pathologist in a topographic order. Histological sections were prepared according to standard procedures of pathology.

FISH was performed on interphase nuclei of 4-µm sections of formalin-fixed and paraffin-embedded (FFPE) tissue. Deparaffinised slides were subjected to several pretreatment steps. They were incubated in 0.2 M HCl for 20 min, followed by washing, and 80°C heat pretreatment in pretreatment-solution (Abbott, Wiesbaden, Germany) for 30 min and further washing. Next, the tissue was digested for 1.5 h with protease, slides were washed again, and fixed in 4% buffered formalin. After washing, dehydration and air drying, an appropriate amount of fluorochrome-labeled probe mixture was applied to the tissue section. Sample DNA and probe mixture were co-denatured in a thermobrite hybridiser (Abbott) at 85°C for 10 min, cooled down to 37°C, and hybridised overnight in a humidified hybridisation chamber. After stringent washing and dehydration and air drying, the sections were counterstained using fluorescence mounting media containing (4′,6-diamidino-2-phenylindole) (DAPI). FISH slides were evaluated with a fluorescence microscope (DM 5500; Leica Microsystems, Wetzlar, Germany) using appropriate filter sets. Orange (centromeric region) and green (target gene) signals were counted in 60 non-overlapping tumor cell nuclei from three tumor areas. Dual color FISH probes were derived from ZytoVision (Bremerhaven, Germany).

Six 10-µm sections of 127 FFPE esophageal tumor biopsies were used for macrodissection and subsequent DNA extraction as previously described (18). Briefly, for genomic DNA extraction, the lesional areas were marked on a haematoxylin and eosin (H&E) slide by a senior pathologist. After deparaffinisation, tumor tissue was macrodissected from the unstained slides and the tissue was lysed with proteinase K overnight. Subsequently, DNA purification was carried out with the BioRobot M48 robotic workstation and the corresponding MagAttract DNA Mini M48 kit (Qiagen, Hilden, Germany) following the manufacturer's protocol.

The TP53 mutation status was evaluated by data analyses available from a comprehensive next generation sequencing approach which was performed on 68 ECs from a total of 127 tumors. Library construction on a hot spot cancer gene panel (Life Technologies, Darmstadt, Germany) and ultra-deep sequencing on the MiSeq platform (Illumina, San Diego, CA, USA) were performed according to Grünewald et al (19).

Evaluation of the MDM2 gene copy number variation (CNV) of the esophageal tumors was performed by quantitative PCR (qPCR). DNA from human embryonic kidney epithelial cells (HEK293) and DNA from 14 macrodissected, non-tumorous colon FPPE tissues were taken as non-MDM2-amplified reference DNA. Primer and probes for CNV assays of MDM2, TERT and RNase P were supplied by Applied Biosystems Life Technologies (Darmstadt, Germany).

Real-time PCR was carried out in triplicate, following the manufacturer's protocol. Copy numbers were then interpreted by the ΔΔCt method using the non-tumorous reference panel as calibrator and RNAse P as endogenous control. Evaluation of the RNAse P gene as the best endogenous control gene, demonstrating stable copy numbers of two alleles per cell in different reference panels (data not shown), was determined by standard curves and analysis of different non-amplified reference DNA.

FISH analysis of gene amplification is laborious, costly and time-consuming. Therefore, we macrodissected the tumor areas from 146 EC specimens and extracted DNA, in order to study MDM2 CNVs by qPCR. DNA qualtity of 127 ECs (Table I) was sufficient to perform qPCR. Normalization of MDM2 gene quantification was established, using RNAse P gene as a calibrator. Determination of RNAase P values in normal tissues ascertained normalisation to two copies per cell (data not shown).

Then, we screened the 127 ECs (Table I) for MDM2 gene amplification. Fig. 1 demonstrates that eight EC (6.3%) samples showed an MDM2 copy number of 4 and higher. These tumors were then studied by FISH showing tumor areas with high MDM2 gene clusters (Fig. 2) and increased MDM2 protein expression in the nuclei of EC tumor cells.

Importantly, in some samples (n=15, 11.8%) we found mean values of more than three MDM2 gene copies per cell by qPCR, indicating that these ECs might be also positve for MDM2 gene amplification (shown in area B of Fig. 1). Therefore, we studied the tumor histology and the MDM2 protein expression of these ECs. Furthermore, gene amplification was analysed by FISH. In most cases, no or only very low gene amplification was shown by FISH and importantly, we did not observe the typical cluster-like amplification pattern of MDM2-positive tumor cells. In addition, ECs with slightly elevated MDM2 copy numbers showed moderately enhanced MDM2 protein expression by immunohistology (Fig. 3).

However, a slightly increased MDM2 copy number by PCR (shown in area B of Fig. 1) might also be due to low tumor cell content in the macrodissected EC area. Indeed, in some case with an MDM copy number >3 (Fig. 2B) only a few tumor cells were observed, that were scattered over the tumorous area, embedded by a high proportion of non-tumorous stroma cells. These cells clearly showed MDM2 gene amplification by a cluster-like pattern of the MDM2 FISH signals (Fig. 4).

From a total of 127 ECs, we identified by qPCR 8 tumors (6.3%) that carry highly elevated numbers of the MDM2 gene (Table II and Fig. 2). There was no prevalence of MDM2 gene amplification in AC or SCC (Table II).

Treatment with MDM2 inhibitors is a novel important therapeutical option, but depends on the TP53 wild-type form. Therefore, we searched the NGS data of the EC cohort and analysed whether tumors with MDM2 amplification and overexpression harbor the mutated or wild-type TP53. From a total of 68 ECs, NGS data were available and revealed that 63% of the ECs had a pathogenic mutation in TP53. From the 8 samples, which were positive for MDM2 amplification, 4 samples carried the wild-type and 4 were shown to carry the TP53 mutants.