Influence of the HER receptor ligand system on sensitivity to cetuximab and trastuzumab in gastric cancer cell lines

The cell lines AGS, Hs746T, KATOIII, LMSU, MKN1, MKN28 and MKN45 were obtained and cultured as described previously (Heindl et al. 2012; Kneissl et al. 2012). The cell lines GSU, H111TC, HGC-27 and MKN7 were provided by the Cell Bank RIKEN BioResource Center (Tsukuba, Japan), and the identity of the cell lines was guaranteed by the provider. GSU, H111TC and MKN7 cells were grown in RPMI-1640 medium (Invitrogen/Gibco, Darmstadt, Germany), and HGC-27 cells were cultured in Eagle’s minimum essential medium (MEM, Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany). Both media were supplemented with 10% foetal bovine serum Sera Plus (PAN Biotech, Aidenbach, Germany) and penicillin–streptomycin (PAA Laboratories, Pasching, Austria; 100 international units (IU)/ml, 100 μg/ml); in addition, RPMI-1640 was supplemented with 2 mM l-glutamine (Invitrogen/Gibco). General cultivation conditions and routine mycoplasma testing as well as cell line validation were performed as described previously (Heindl et al. 2012; Kneissl et al. 2012).

For Western blot analysis, the following antibodies were used: anti-EGFR (Cell Signaling, Leiden, NL, #2232), anti-pEGFR (Y1068) (Invitrogen, #44788G), anti-HER2 (Cell Signaling, #2165), anti-pHER2 (Y1248) (Cell Signaling, #2247), anti-HER3 (Cell Signaling, #4754), anti-pHER3 (Y1222) (Cell Signaling, #4784), anti-HER4 (Cell Signaling, #4795), anti-pHER4 (Y1284) (Cell Signaling, #4757), anti-TACE (Cell Signaling, #6978), anti-β-actin (Sigma-Aldrich, #A1978), anti-α-tubulin (Sigma-Aldrich, #T9026), anti-rabbit IgG (Cell Signaling, #7074) and anti-mouse IgG (GE Healthcare, Munich, Germany, #NA931).

The following monoclonal therapeutic antibodies were used: cetuximab (Erbitux™, Merck Serono, Darmstadt, Germany), trastuzumab (Herceptin™, Roche, Penzberg, Germany) and isotype control (Southern Biotech, Birmingham, USA, #0151K-14). The corresponding solvent controls were as follows: solvent control isotype: 1 × PBS; solvent control trastuzumab: 3.36 mg l-histidine HCl, 2.16 mg l-histidine, 136.2 mg trehalose dihydrate, 0.6 mg polysorbate 20, dissolved in 7.2 ml sterile water (http://​www.​ema.​europa.​eu/​docs/​en_​GB/​document_​library/​EPAR_​-_​Scientific_​Discussion/​human/​000278/​WC500049816.​pdf).

The solvent control for cetuximab was described previously (Heindl et al. 2012).

Recombinant ligands were obtained as follows: human AREG (R&D systems, Minneapolis, USA, #262-AR-100), human HB-EGF (Pelobiotech, Planegg, Germany; #PB-Z3051) and human EGF (Sigma-Aldrich, #E9644).

Chemotherapeutics were obtained as follows: 5-fluorouracil (Sigma-Aldrich, #F6627), cisplatin (Sigma-Aldrich, #P4394).

To assess the effects of the treatments on cell proliferation, the WST-1 cell proliferation assay was used according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany; #11644807001). All samples were analysed in triplicate. Cells were seeded at densities between 0.5 × 10³ and 5 × 10³ cells per well in 80 µl culture medium and allowed to settle for 24 h. The following day, cetuximab and/or trastuzumab and/or recombinant human AREG, EGF or HB-EGF and/or chemotherapeutics were added to a final volume of 100 µl per well. The isotype control was added to a final concentration of 100 µg/ml for cetuximab and 40 µg/ml for trastuzumab. Due to the high volume needed, we investigated the effect of the isotype solvent as well. The applied volume for the cetuximab solvent corresponded to 100 µg/ml cetuximab, and the trastuzumab solvent corresponded to 40 µg/ml trastuzumab. Assays using cetuximab were incubated for 48 h, and assays with trastuzumab were incubated for 72 h. Experiments with concomitant treatment of cetuximab and trastuzumab were incubated for 72 h as well. After this incubation period, pre-warmed WST-1 reagent was added. The absorbance of the samples was measured after an incubation period between 30 min and 2 h, depending on the cell line. An Asys Expert Plus microplate reader was used for measurements (Biochrom, Berlin, Germany).

For extraction of genomic DNA from GSU and H111TC cells, the DNeasy kit was used according to manufacturer’s instructions (Qiagen, Hilden, Germany; #69504). The DNA concentration was measured using the QuBit 2.0 DNA high sensitivity kit (Thermo Fisher Scientific, Waltham, USA, #Q32854). Furthermore, DNA sequencing grade quality was determined by a qPCR assay (RNAse P assay, Thermo Fisher Scientific, Waltham, USA, #4316831) as described previously (Endris et al. 2013).

For library preparation, the multiplex PCR-based Ion Torrent AmpliSeq™ technology (Thermo Fisher Scientific, Waltham, USA), together with the Cancer Hotspot Panel (CHPv2; Thermo Fisher Scientific, #4475346), was used as described previously (Endris et al. 2013; Stenzinger et al. 2014).

Amplicon library preparation was performed with the Ion AmpliSeq Library Kit v2.0 (Thermo Fisher Scientific, #4480442). For mutation analysis, the CHPv2 panel, which consists of one primer pool yielding 207 amplicons covering hot spot regions of 50 known cancer-related genes, was employed. For amplification, approximately 10 ng of DNA, as determined by qPCR assay, was used. Briefly, the DNA was mixed with the primer pool and the AmpliSeq HiFi Master Mix in a 20-µl reaction volume and transferred to a PCR cycler (Biometra, Göttingen, Germany). After the end of the PCR, amplicons were partially digested using FuPa reagent, followed by the ligation of barcoded sequencing adapters (Ion Xpress Barcode Adapters, Thermo Fisher Scientific, #4474517). The final library was purified using AMPure XP magnetic beads (Beckman Coulter, Krefeld, Germany, # A63880) and quantified using qPCR (Ion Library Quantitation Kit, Thermo Fisher Scientific, #4468802) on a StepOnePlus qPCR machine (Thermo Fisher Scientific, Waltham, USA). The individual libraries were diluted to a final concentration of 100 pM. All libraries were pooled and processed for library amplification on Ion Spheres using Ion PGM™ Template OT2 200 Kit (Thermo Fisher Scientific, #4480974). Unenriched libraries were quality controlled using an Ion Sphere quality control measurement on a QuBit instrument. After library enrichment (Ion OneTouch ES, Thermo Fisher Scientific), the library was processed for sequencing using the Ion PGM™ Sequencing 200 Kit v2 chemistry (Thermo Fisher Scientific, #4482006) and the barcoded libraries were loaded onto a 318v2 chip.

Raw sequencing data were processed using the implemented Torrent Suite software (version 4.4.3) and aligned with the human genome (version hg19) using the TMAP algorithm. For DNA mutation analysis, the aligned reads were processed using the built-in Variant Caller plugin (version 4.4.3). Variant annotation was performed using a custom-build variant annotation pipeline in the CLC Genomics Workbench (version 8.0.2). For visualization of sequencing and fusion reads, the Integrative Genomic Browser (IGV, http://​www.​broadinstitute.​org/​igv/​) was used. We used the COSMIC (catalogue of somatic mutations in cancer) database (Forbes et al. 2015; Sherry et al. 2001) to identify already known somatic mutations and mutation types, respectively.

Copy number variations (CNVs; amplifications and deletions) were identified using the coverage data summary for each sample and each amplicon generated by the Torrent Suite software. Detection of CNVs was performed according to Endris et al. (2013). In brief, gene amplifications and/or deletions were determined by a simple algorithm using the number of reads of each individual amplicon in the sequenced sample: (i) the number of reads of each individual pool amplicon was divided by the total number of sequencing reads of the respective sample = (reads amplicon x/total reads) = normalized amplicon read depth value (NARD); (ii) the NARD was multiplied by the total number of amplicons (e.g. lung cancer panel = 140 amplicons; NARD (reads amplicon x/total reads) × 140); (iii) the median normalized amplicon read depth (MNARD) was determined for all samples = median (NARDSample1:NARDSampleX), reflecting the typical amplification efficiency of each individual amplicon in the pool; and iv) the standard deviation (SD) was determined from the median value. Amplifications are considered true if the NARDs of all amplicons covering a gene differ by >2 SD from the median value. However, deletions are considered true if the SD of all amplicons covering a gene is <0.5.

Array-comparative genomic hybridization (aCGH) was performed as previously described (Juskevicius et al. 2016; Ruiz et al. 2011), with minor modifications. In brief, 500 ng of sample DNA (cell line DNA of GSU or H111TC) and equal amounts of female reference genomic DNA (Promega, Madison, WI, USA) were digested with DNaseI to a size range of 200–500 bp. Subsequent labelling of sample and reference DNA with Cy3-dUTP and Cy5-dUTP, respectively, was performed with the BioPrime^® Array CGH Genomic Labeling System (Invitrogen, Carlsbad, CA, USA). The success of labelling was assessed by quantifying the specific activities of the incorporated dyes with a Nanodrop (Thermo Fischer Scientific, Waltham, MA, USA). Reference and sample DNA were mixed and hybridized to 180 k CGH arrays (Agilent Technologies, Santa Clara, CA, USA) for 24 h in a rotating oven at 67 °C. Microarray slides were scanned with the Agilent 2565C DNA scanner, and images were analysed with Agilent’s Feature Extraction using default settings. Feature extracted array CGH data were evaluated using Agilent’s CytoGenomics software v3.0.1.1. Aberrations were called with the aberration detection algorithm ADM2 set to a threshold of 12.0, with fuzzy zero and GC-content (window size: 2 kb) correction. A minimum of three probes were necessary to call an aberration.

For detection of human AREG, HB-EGF and TGFα, DuoSet ELISAs were used (R&D systems, Minneapolis, USA; #DY262, DY259, DY239) according to the manufacturer’s instruction. All samples were analysed in duplicates.

To determine the levels of secreted ligand (AREG, HB-EGF, TGFα), 1 × 10⁶ cells were seeded into 10 ml medium and incubated for 24 h. Next, conditioned cell culture medium was harvested and centrifuged at maximum speed (4 °C) for 10 min to remove cell debris. Aliquots were stored at −80 °C.

For analysis of the effect of trastuzumab treatment in combination with an exogenous ligand application on ligand secretion, 2 × 10⁵ cells (MKN45, GSU) were seeded into 2 ml medium. Cells were allowed to settle overnight. The following day, the cells were treated with medium containing 10 µg/ml trastuzumab and/or 15 ng/ml AREG or 0.1 ng/ml EGF or 0.4 ng/ml HB-EGF for 6 h. The conditioned medium was harvested and stored as mentioned above.

The effect of extensive trastuzumab and cetuximab treatment was analysed by seeding 2 × 10⁵ cells into 6 ml medium. After 24 h, 10 µg/ml cetuximab or trastuzumab was added. The cells were incubated for 8 days with medium changes every 3–4 days. Then, the cells were trypsinized and 2 × 10⁵ cells were seeded into 2 ml medium containing 10 µg/ml cetuximab or trastuzumab. Conditioned medium was harvested after 24 h as mentioned above.

For analysis of the effect of trastuzumab treatment in combination with an exogenous ligand on protein levels, 2 × 10⁵ cells (MKN45, GSU) were seeded into 2 ml medium. Cells were allowed to settle overnight. The following day, the cells were treated with medium containing 10 µg/ml trastuzumab and/or 15 ng/ml AREG, 0.1 ng/ml EGF or 0.4 ng/ml HB-EGF for 6 h. The effect of extensive trastuzumab and cetuximab treatment was analysed by seeding 2 × 10⁵ cells into 6 ml medium. After 24 h, 10 µg/ml cetuximab or trastuzumab was added. The cells were incubated for 8 days with medium changes every 3–4 days. Cell lysates were prepared as described previously (Bremm et al. 2008). Western blot analysis was performed using a standard protocol described previously. The antibodies were used in the following dilutions: anti-EGFR, anti-HER2, anti-pHER2, anti-HER3, anti-pHER3, anti-HER4 and anti-pHER4: 1:1000 in 5% BSA-TBS-T (w/v); anti-pEGFR: 1:2000 in 5% milk-TBS-T (w/v); anti-TACE: 1:3000 in 5% milk-TBS-T (w/v); anti-β-actin, anti-α-tubulin, anti-mouse IgG: 1:10,000 in 5% milk-TBS-T (w/v); anti-rabbit IgG: 1:2000 in TBS-T. Signal detection was performed using an enhanced chemiluminescence reaction. The signals were quantified by densitometric measurement via ImageJ software 1.42q (National Institute of Health, MD, USA).

A PubMed search was performed for relevant gastric cancer-based literature about the HER receptor ligands AREG, EGF, HB-EGF and TGFα (date: 6 May 2015). The search was performed using the following terms: amphiregulin gastric cancer, EGF gastric cancer, HB-EGF gastric cancer, TGFα gastric cancer.

Inclusion criteria were as follows: (1) published as an original article, (2) published in the English language, (3) full-text access, (4) examined ligand mRNA levels, ligand protein levels or pro-ligand protein levels in (5) tissue, serum, gastric juice and other body fluids of gastric cancer patients. Studies dealing only with tumours of the gastroesophageal junction (GEJ) were excluded as well as studies based on less than 10 patients and studies with an unclear number of tumour samples. Furthermore, studies concerning gene polymorphisms in ligand genes and their association with gastric cancer risk were excluded. Additionally, all studies regarding only co-expression of HER ligands in combination with other proteins were not included. The studies were screened for relevant information regarding the expression rates of the ligands and their correlation to clinicopathologic features.

All analyses presented herein were performed in at least three independent experiments. Statistical analyses were calculated using IBM SPSS Statistics 22 (IBM, Armonk, NY, USA). The results are shown as the mean value ± standard deviation (SD). For comparing pairs of different treatment conditions, the two-sided Welch t test was used. When values were compared to a reference value (=100%), we used the one-sample t test. p values ≤0.05 are indicated by * and ≤0.01 by **. The authors will provide all statistical analyses on request.

In this study, a panel of 11 gastric cancer cell lines was used. In addition to the data we published recently (Heindl et al. 2012; Kneissl et al. 2012), we identified the cell lines GSU, H111TC and MKN7 as cetuximab sensitive (Table 1, Online Resource 1). Following treatment with 10 µg/ml cetuximab, GSU cells displayed a decrease in cell proliferation to 68.10% in comparison with the untreated control (p = 0.04). H111TC cells were less sensitive with growth rates of 82.30% after treatment with 10 µg/ml cetuximab (p = 0.002) and a maximum inhibition to 80.16% after application of 200 µg/ml (p < 0.001). MKN7 cells showed only a minor sensitivity with a maximum decrease of cell proliferation to 83.63% following treatment with 100 µg/ml cetuximab (p = 0.001). HGC-27 cells were completely resistant to cetuximab.

Additionally, we investigated the trastuzumab sensitivity of the cell lines via cell proliferation assay (Fig. 1). We identified two cell lines as trastuzumab sensitive: GSU and H111TC. Both displayed a significant decrease in cell proliferation after application of the therapeutic. For GSU, 0.1 µg/ml was already sufficient to inhibit the proliferation rate to 80.10% (p = 0.009), while 40 µg/ml caused an inhibition to 72.98% (p = 0.002). H111TC displayed a less sensitive phenotype with a maximum inhibition to 80.03 and 80.31% after application of 20 µg/ml and 40 µg/ml trastuzumab (p = 0.024; p = 0.02), respectively. For the other cell lines, no significant inhibition of cell proliferation was observed.

Among the numerous genetic alterations which have been connected to resistance to cetuximab, activating mutations in the KRAS gene are especially important in colorectal cancer regimens (Karapetis et al. 2008; Lievre et al. 2006). Regarding trastuzumab therapy, activating mutations in the PIK3CA gene have been described to be associated with a poorer therapy outcome in breast cancer patients (Berns et al. 2007; Majewski et al. 2015). Additionally, the existence of HER2 amplifications and the resulting consecutive overexpression of the protein are crucial factors in the response predictions for trastuzumab-based regimens in patients with gastric cancer (Okines et al. 2010). Therefore, we collected information about these three genetic loci in our panel of gastric cancer cell lines. We were able to extract some data from the literature and completed the missing data for the majority of the cell lines with our own analyses (Table 1).

Interestingly, we found that the cetuximab-sensitive cell line GSU harbours an activating KRAS mutation in exon 2 (G12D), and MKN1 cells, which are cetuximab sensitive, were described as KRAS amplified (Mita et al. 2009). However, activating PIK3CA mutations have only been described for trastuzumab-insensitive cell lines (MKN1, HGC-27, AGS, MKN45) (Mita et al. 2009; Zhou et al. 2011), while we found both sensitive cell lines, GSU and H111TC, to be wild type. Regarding the HER2-amplification status, the picture was less consistent as we found a suspected HER2 amplification only for the cell line H111TC by next-generation sequencing (Online Resource 2) which was confirmed by array-comparative genomic hybridization (ERBB2/CEP7 ratio > 2.0) (Online Resource 3, 4).

In our cells, trastuzumab monotreatment showed only a moderate effect on cell proliferation. Therefore, we decided to investigate the effect of a concomitant application of cetuximab or chemotherapeutics (5-fluorouracil [5-FU] and cisplatin) to trastuzumab.

Although GSU cells are sensitive to trastuzumab as well as to cetuximab, a combination of both monoclonal antibodies failed to show any enhancing effect in comparison with the monotreatment (Fig. 2a). However, the addition of 5-FU and cisplatin to trastuzumab led to significantly stronger growth inhibition in GSU cells than trastuzumab or chemotherapy alone (Fig. 2b). This effect was not observed in the trastuzumab-insensitive cell line MKN45, although this cell line reacted to the monotreatment with chemotherapeutics with a significant growth inhibition (p = 0.02).

To determine the effect of an extended application of trastuzumab and cetuximab on the expression and activation of the HER receptors, we treated all cell lines with the therapeutics for 8 days and measured the expression of EGFR, pEGFR, HER2, pHER2, HER3, pHER3, HER4 and pHER4 (Figs. 3, 4, Online Resource 5, 6). The basal expression rate of EGFR, pEGFR, HER2, pHER2, HER3 and pHER3 varied highly between the cell lines. The highest EGFR levels were expressed by MKN1 cells; AGS, GSU, Hs746T, KATOIII, MKN7 and MKN28 cells displayed medium expression levels, and in MKN45, LMSU, H111TC and HGC-27 cells, EGFR was hardly detectable. In contrast to these findings, pEGFR was expressed at the highest levels by KATOIII and MKN45 cells, a medium expression rate was detected in GSU, H111TC, Hs746T and MKN7 cells, and pEGFR was barely detectable in AGS, HGC-27, LMSU, MKN1 and MKN28 cells. For AGS, Hs746T, KATOIII, LMSU, MKN1, MKN28 and MKN45, these findings support our recently published results (Heindl et al. 2012; Kneissl et al. 2012). Regarding HER2, the highest levels were expressed by MKN7 cells, medium levels were expressed by GSU, H111TC, HGC-27, KATOIII, LMSU and MKN1 cells, and HER2 expression was hardly detectable in AGS, Hs746T, MKN28 and MKN45 cells. We were able to detect medium levels of pHER2 in GSU, H111TC, Hs746T KATOIII, MKN7 and MKN45 cells, while in all other cell lines, the protein was hardly detectable. All cell lines, with the exception of LMSU (no expression) and GSU (high expression), expressed HER3 at median levels; however, we were only able to regularly detect pHER3 above the background level in KATOIII cells. pHER4 could not be detected in any cell line, and only HGC-27 cells displayed a medium expression of HER4. Regarding the effect of 8 days of cetuximab treatment on the expression levels, we found no obvious patterns for EGFR, pEGFR, HER3 and pHER3. However, densitometric measurement revealed an increase in the expression of HER2 after treatment for 8 of 11 cell lines (significant for H111TC and MKN7 cells; p = 0.041; p = 0.029). pHER2 expression was enhanced in 10 out of 11 cell lines, but this effect was only significant for MKN7 (p = 0.028). Regarding the effect of 8 days of trastuzumab treatment, relevant patterns were only observed for HER2. H111TC, HGC-27, MKN1 and MKN28 all displayed a significant decrease in HER2 expression after trastuzumab treatment (p = 0.003; 0.01; 0.038; 0.015, respectively). Furthermore, pHER3 levels significantly decreased in Hs746T and MKN7 cells. Additionally, we investigated the effect of the trastuzumab and cetuximab treatment on the expression of the protein TACE which is responsible for the cleavage of the membrane-bound ligand precursors into the soluble ligand [for review: (Mochizuki and Okada 2007; Sternlicht and Sunnarborg 2008)]. We detected only for GSU cells after trastuzumab treatment a significant decrease indicating only a minor influence of cetuximab or trastuzumab treatment on TACE expression in gastric cancer cell lines.

We analysed the level of HB-EGF and TGFα in our panel of gastric cancer cell lines. For most cell lines, AREG secretion had already been determined in a previous study (Table 1; Kneissl et al. 2012). In addition, we found that GSU, H111TC and MKN7 cells secrete high amounts of AREG to the medium, while the AREG secretion of HGC-27 was barely detectable (Table 1, Online Resource 7).

TGFα levels in the conditioned medium of the cell cultures were analysed, and a broad range was detected (Fig. 5a). TGFα was hardly detectable with values below 5 pg/ml in H111TC, HGC-27, LMSU and MKN45 cells. The cell lines AGS, GSU, Hs746T and MKN1 secreted medium levels of TGFα in a range between 5 and 10 pg/ml in the medium. The highest concentrations were detected for KATOIII, MKN7 and MKN28 cells.

HB-EGF was secreted in different concentrations; however, for the majority of the cell lines, HB-EGF levels were below 20 pg/ml. Substantially higher concentrations were only secreted by Hs746T, MKN7 and MKN28 cells (Fig. 5b).

To investigate the effect of prolonged trastuzumab treatment on HER receptor ligand secretion, we treated GSU cells for 8 days with 10 µg/ml trastuzumab and measured the level of AREG, HB-EGF and TGFα in the cell culture supernatant. Although we did not observe any significant effects on ligand secretion, we found a decreasing trend in AREG secretion (p = 0.099). No relevant differences regarding the levels of HB-EGF and TGFα were detected (Fig. 6a).

In contrast, we observed after 8 days of cetuximab treatment an increase in HB-EGF and TGFα levels in MKN1 and Hs746T cells, although these effects were non-significant. Similar to trastuzumab treatment, application of cetuximab caused a significant decrease in AREG secretion in the cetuximab-sensitive cell line MKN1 (p = 0.043), while AREG levels in Hs746T were not altered (Fig. 6b).

To investigate the effects of exogenous ligands on the trastuzumab sensitivity of gastric cancer cells, we performed a concomitant treatment of gastric cancer cell lines with AREG, EGF or HB-EGF and trastuzumab. We then measured the influence of this treatment on the metabolic activity of the cells via the WST-1 cell proliferation assay.

Ligand application had no effect on the proliferative activity of the trastuzumab-insensitive cell lines KATOIII and MKN45, as no differences between the treatments could be observed. In contrast, in the trastuzumab-sensitive cell lines GSU and H111TC, a combination of the ligands in addition to trastuzumab influenced the trastuzumab sensitivity in different ways. In GSU cells, trastuzumab application (20 µg/ml) inhibited the proliferative activity to 82.7% compared to the untreated control (p = 0.017). EGF and AREG were ineffective in rescuing the cells from growth inhibition, as there was no significant difference compared with the control. In contrast, HB-EGF completely neutralized the inhibitory effect of trastuzumab. For H111TC cells, HB-EGF application partially rescued the cells from trastuzumab inhibition; however, this effect was not significant. Additionally, we observed enhanced trastuzumab sensitivity of the cells after application of exogenous AREG (Fig. 7; Table 2).

Western blots were performed to investigate the effects of parallel treatment with trastuzumab and AREG, EGF, HB-EGF on pEGFR and pHER2 levels. For these experiments, a 6-h treatment was chosen. While no effect on the trastuzumab-insensitive cell line MKN45 could be observed, GSU showed an increase in pEGFR levels after the trastuzumab treatment in comparison with the untreated control (p = 0.038). Interestingly, concomitant application of either ligand suppressed this effect. However, only for HB-EGF was this suppression significant (p = 0.05). Regarding pHER2, similar, but non-significant patterns were observed (Fig. 8, Online Resource 8).

In addition, we measured the levels of the ligands AREG, HB-EGF and TGFα in the conditioned medium of GSU and MKN45 cells, following treatment for 6 h with trastuzumab and/or exogenous ligand (AREG, EGF, HB-EGF). Levels of HB-EGF and AREG showed no significant change with any of the treatments, but application of each ligand caused a significantly increased secretion of TGFα in GSU cells. Regarding this effect, HB-EGF was more effective than EGF, and EGF was more effective than AREG. Interestingly, in the trastuzumab-insensitive cell line MKN45, a significant decrease of TGFα was observed after treatment with AREG/trastuzumab and EGF (Fig. 9).

We investigated the effects of different concentrations of AREG, EGF and HB-EGF on the cetuximab sensitivity of the sensitive cell line MKN1 and the resistant cell line Hs746T. We found both, HB-EGF and EGF, but not AREG, to be effective in rescuing MKN1 from cetuximab inhibition. Surprisingly, we detected a minor inhibition in the cetuximab-resistant cell line Hs746T by cetuximab application; however, exogenous ligand application had no additional effect (Online Resource 9).

To investigate the relevance of HER receptor ligands in gastric tumours, a literature research was performed. As shown in Table 3, the expression of AREG, EGF, HB-EGF and TGFα in gastric tumours and body fluids of gastric cancer patients has been analysed in multiple studies. The majority of the articles investigated the expression of EGF and TGFα.

EGF protein expression was found in up to 88% of the gastric tumour samples, with most studies reporting EGF positivity in 29–58% of the cases (Aoyagi et al. 2001; Borlinghaus et al. 1993; Docea et al. 2013; Hirayama et al. 1992; Livingstone et al. 1995; Oda et al. 1990; Onda et al. 1990; Pryczynicz et al. 2009; Sugiyama et al. 1989; Tahara et al. 1986; Yasui et al. 1988; Yoshiyuki et al. 1990). EGF was detected in the serum, plasma, urine and gastric juice of patients with gastric cancer, while analysis of the malignant ascites revealed only very low EGF levels (Chuang et al. 1994; Dias et al. 2011; Dragovich et al. 2006; Han et al. 2009; Park do et al. 2014; Yasumoto et al. 2011; Zhang et al. 2014). Several studies indicated an association with advanced disease, metastatic disease and poor prognosis (Czyzewska et al. 2009; Hirayama et al. 1992; Onda et al. 1990; Tahara et al. 1986; Yasui et al. 1988), while in early gastric cancers, per trend a smaller percentage of EGF-positive tumours was reported (Aoyagi et al. 2001; Hirayama et al. 1992; Onda et al. 1990; Tahara et al. 1986; Yasui et al. 1988). Regarding the predictive value of EGF expression for EGFR-targeted therapies, the results are completely contradictory; in patients treated with cetuximab combined with modified FOLFOX6, low EGF levels were associated with a higher response rate (Han et al. 2009). In contrast, patients with high EGF serum levels treated with cetuximab, cisplatin and capecitabine displayed a longer overall survival (Zhang et al. 2014). In addition, EGF plasma levels were not found to be predictive for therapy response in patients treated with erlotinib (Dragovich et al. 2006).

Similarly inconsistent findings concerning EGFR-targeted therapies were reported for TGFα (Dragovich et al. 2006; Han et al. 2009; Zhang et al. 2014). Expression of the TGFα protein was reported in most studies in between 35 and 74% of gastric tumours (Celikel et al. 2007; Dragovich et al. 2006; Espinoza et al. 2004; Konturek et al. 2001; Livingstone et al. 1995; Muller and Borchard 1992; Saeki et al. 1994). Regarding the general prognostic value of TGFα, data are quite inconsistent; high TGFα levels in the tumour were correlated with lymph node metastasis, poor overall survival, advanced TNM stage and tumour size in some studies (Aoyagi et al. 2001; Celikel et al. 2007; Espinoza et al. 2004; Fanelli et al. 2012). On the other hand, no correlation with clinicopathologic features and prognosis was reported in another study (Muller and Borchard 1992). Furthermore, serum TGFα levels showed no correlation with clinicopathologic characteristics (Choi et al. 1999).

Only a few studies have investigated the expression of AREG in gastric cancer. Positive protein staining was reported for 34.8% (pro-AREG) and 51% (AREG) of the tumours (Saeki et al. 1994; Yoshida et al. 2012). AREG mRNA was detected in malignant as well as in non-malignant tissue (Kitadai et al. 1993; Naef et al. 1996; Nielsen et al. 2014). However, an increase of AREG mRNA expression in tumour tissue was reported twice (Kitadai et al. 1993; Nielsen et al. 2014). One study found no correlation between AREG mRNA expression and staging or histological tumour types, and another study reported a significantly shorter median survival in patients with pro-AREG-positive tumours (Kitadai et al. 1993; Yoshida et al. 2012). In addition, high AREG concentrations were found in the malignant ascites of gastric cancer patients (Yasumoto et al. 2011). AREG serum levels were not predictive for response rate, median time to progression or median overall survival (Han et al. 2009).

Regarding HB-EGF, data are highly consistent and indicate an association with advanced disease: levels of soluble HB-EGF were found to be elevated in the serum of advanced gastric cancer patients (Chung et al. 2015) as well as in malignant ascites of gastric cancer patients (Yasumoto et al. 2011). Furthermore, increased expression of proHB-EGF and of the cytoplasmic domain of HB-EGF in advanced tumour stages was reported (Murayama et al. 2002; Shimura et al. 2012). Interestingly, there is additional evidence that HB-EGF might be a resistance factor against 5-FU- and cisplatin-based chemotherapies (Suganuma et al. 2003).