The present study was conducted on a previously reported group of patients (26) that was enriched with additional cases to achieve a total study population of 145 cases (all female). The eligibility criteria were as follows: i) Treatment with trastuzumab for histologically confirmed MBC; ii) adequate clinical data on patient history, demographics, tumour characteristics, treatment details (i.e., drug dosages, schedule of administration and serious toxicities) and clinical outcome; and iii) adequate tumour tissue available for biological marker evaluation.

Formalin-fixed paraffin-embedded tumour tissue samples were retrospectively collected from patients treated with trastuzumab-based regimens in the metastatic setting, as previously described in detail (26–28). All tumours were characterised by local pathologists as HER2-positive based on American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) criteria current at that time (29). Consequently, all patients received trastuzumab as part of their treatment. All tumour samples were centrally re-evaluated by IHC for oestrogen receptors (ER), progesterone receptors (PgR), HER2 and the expression of the proliferation marker Ki67. Additionally, HER2 amplification status was assessed using the FISH method, as later described.

The translational research protocol was approved by the Bioethics Committee of the Aristotle University of Thessaloniki School of Medicine (Protocol #4283; January 14, 2008; Thessaloniki, Greece) under the general title ‘Investigation of major mechanisms of resistance to treatment with trastuzumab in patients with metastatic breast cancer’. All patients included in the study in 2005 and later provided written informed consent for the provision of biological material for future research studies before receiving any treatment. A waiver of consent was obtained from the Bioethics Committee for patients included in the study before 2005.

Representative haematoxylin-eosin-stained 2-µm sections from the tissue blocks were reviewed by a pathologist. A total of 17 TMA blocks were constructed from the 145 eligible cases using a manual tissue microarrayer (Beecher Instruments), as previously described (26). For the construction of the TMA blocks, two core samples (1.5 mm in diameter) were obtained from representative regions of each tumour in the donor blocks. All IHC and FISH markers were assessed on the TMA sections. The cases not represented, which had damaged or inadequate cores on the TMA sections, were re-cut from the original blocks if still available. These sections were used for protein and gene analyses as previously described (26–28,30). Further method details are provided in Data S1.

Serial 2.5-µm-thick TMA sections or whole tissue sections were stained for ER (clone 6F11; cat. no. NCL-L-ER-6F11; Leica Microsystems, Ltd.), PgR (clone 1A6; cat. no. NCL-L-PGR; Leica Microsystems, Ltd.), HER2 (polyclonal Ab; cat. no. A0485; Dako; Agilent Technologies, Inc.), Ki67 (clone MIB-1; cat. no. M7240; Dako; Agilent Technologies, Inc.), phosphorylated mTOR at serine 2448 (p-mTOR; clone 49F9; cat. no. 2976; Cell Signaling Technology, Inc.; dilution, 1:30; 20 min) and stPTEN (clone 6H2.1; cat. no. M3627; Dako; Cell Signaling Technology, Inc.; dilution, 1:200; 30 min), using the Bond Max™ autostainer (Leica Microsystems, Ltd.) as previously described in detail (31) (Table SI). All sections were stained in one run for each antibody and were evaluated by pathologists experienced in breast cancer and blinded to the patient's clinical and survival data. Positive controls were used for all antibodies from known positive breast cancer cases, while negative controls were obtained by omitting the primary antibody as previously described (26–28,30).

All IHC stains were evaluated according to the formerly outlined interpretation (32). Stromal TIL density was assessed on whole H&E sections according to the guidelines from the International TILs Working Group (33) and was analysed as a continuous variable. HER2 protein expression was scored according to the 2007 ASCO/CAP guidelines (29) (scores between 0 and 3+). A positive for HER2 protein expression (IHC 3+) was defined as uniform intense membrane staining of >30% of invasive tumour cells. Further method details are provided in Data S1.

TMA sections or whole tissue sections (5-µm-thick) were used for FISH analysis using the ZytoLight^® SPEC HER2/TOP2A/CEN17 triple colour probe kit for HER2 (code Z-2093; ZytoVision GmbH). FISH was performed according to the manufacturer's protocol with minor modifications (pepsin solution was applied to the specimens and incubated for 11–12 min at 37°C in a humidified chamber) in all cases (i.e., not only the HER2 IHC 2+ cases). Digital images were constructed using software specifically developed for cytogenetics (XCyto-Gen, ver 1.2.11; Alphelys) and evaluated as previously described (26,31). For the assessment of HER2 status, the 2007 ASCO/CAP guidelines (29) were used with the addition of the ≥6-HER2-copies criterion (34), as patients had locally received trastuzumab according to this classification. HER2 status was considered positive if HER2 was amplified by FISH and/or a HER2 score of 3+ was obtained by IHC. Representative IHC and FISH images from invasive breast carcinoma cases are presented in Fig. 1. Further method details are provided in Data S1.

Simultaneous isolation of DNA and RNA from whole or macrodissected 10-µm paraffin sections, the latter in cases with <50% tumour cell content, was performed for the 145 examined tumours with iron oxide beads coated with a nanolayer of silica using the VERSANT Tissue Preparation Reagents kit (Siemens Healthcare Diagnostics) as previously described (35). The nucleic acid extract of each sample was divided into two aliquots. DNase I was then added to one aliquot in order to remove DNA and ensure the presence of pure RNA for downstream mRNA analyses. cDNA synthesis was performed with random primers and SuperScript III Reverse Transcriptase (catalogue number 48190011 and 18080044; Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. cDNAs were assessed in duplicate with quantitative PCR using an ABI7900HT system under default thermal cycling conditions as follows: 50°C for 2 min, 95°C for 10 min, followed by 45 cycles at 95°C for 15 sec and 60°C for 1 min, respectively.

mRNA expression analysis was performed with pre-made exon-spanning TaqMan-MGB assays (Applied Biosystems; Thermo Fisher Scientific, Inc.) targeting the following gene transcripts (data in parentheses refer to assay ID; exon boundary; amplicon length): AREG Hs00950669_m1; 3–4; 66 bp), BTC Hs01101204_m1; 4–5, 5–6; 139 bp), EGF Hs01099999_m1; 19–20, 20–21: 70 bp), EREG ID Hs00914313_m1; 3–4; 65 bp), HBEGF Hs00181813_m1; 3–4; 78 bp), IGFBP4 Hs00181767_m1; 1–2: 91 bp), NRG1 Hs00247620_m1; 2–3: 93 bp), RARA Hs00940446_m1; 4–5, 5–6, 6–7: 68 bp), TGFA Hs00608187_m1; 4–5: 70 bp), TGFB1 Hs00171257_m1; 1–2: 63 bp), THRA Hs00268470_m1; 4–5: 89 bp) (Table I). A TaqMan-MGB expression assay targeting the glucuronidase β (GUSB) gene (Hs99999908_m1; 9–10, 10–11, 11–12; 81 bp) was used as the endogenous reference for relative quantification. The commercially available TaqMan Control Total RNA (cat. no. 4307281; Applied Biosystems; Thermo Fisher Scientific, Inc.) was applied as a positive control for interrun evaluation of the PCR assay efficiency, together with no-template controls. To obtain linear relative quantification (RQ) values, relative expression was assessed as 40-∆Cq, whereby ∆Cq was calculated as (average target Cq) - (average GUSB Cq) from all eligible measurements, as previously described (35). Only samples with average GUSB Cq values <36 and a ∆RQ value for each duplicate sample pair (intrarun variation) of <1 were considered eligible for further analysis in the present study.

Frequencies with the corresponding percentages and medians with range were used to summarize categorical and continuous variables, respectively. Comparisons of categorical data were performed using the χ² or Fisher's exact test (if more appropriate), while the non-parametric Wilcoxon rank-sum test was applied for the comparison of categorical with continuous variables. Associations of the markers of interest were assessed using Spearman's correlations.

The median value of the mRNA expression of the examined markers was used as the optimal cut-off to classify tumours into high- and low-expression groups and assess their prognostic significance and the associations with several clinicopathological characteristics (including age, menopausal status, histological grade, TIL density, PTEN expression and bone metastasis). Additionally, we evaluated the upper and lower quartiles of the mRNA distribution as potential thresholds. Time to progression (TTP) was defined as the time interval between the initiation of trastuzumab-based treatment for advanced disease (with or without parallel administration of chemotherapy or hormonal therapy) and the first documented disease progression. Non-progressors were censored at the date of last follow-up. Survival was also estimated from the initiation of trastuzumab-based therapy until death (from any cause) or last follow-up. Time-to-event distributions were estimated with the Kaplan-Meier method and comparisons of groups were performed with the log-rank test. Cox proportional hazard regression models (univariate and multivariate) were applied to estimate the effect of the examined markers on TTP and survival.

The TTP and survival analyses were conducted separately in the subgroup of HER2-positive and HER2-negative patients (as defined by HER2 central assessment), while the interactions with the ER/PgR status among patients with HER2-positive tumours and with the disease presentation status in the entire study population were also assessed for all examined markers with respect to TTP and survival to detect whether the effect of the marker expression on patients' outcome varied between the subgroups defined by ER/PgR (positive vs. negative) and disease presentation status [relapsed MBC (R-MBC) vs. de novo MBC].

Model choice was performed in multivariate analyses using the following variables in the first step: Menopausal status, performance status, mTOR protein expression, PTEN protein expression and each one of the markers that showed significance in univariate analysis. The final model was selected using backward selection criteria with P<0.10. All tests were two-sided and significance was set at 5%. The SAS software (version 9.3; SAS Institute, Inc.) was used for the statistical analyses.

A total of 145 patients with at least one measurement in the markers of interest, including HER family receptor ligands and other potential prognostic biomarkers, were included in the present study. Among them, 109 (75.2%) had available mRNA data for all 11 markers. Although all patients were found to be HER2-positive when assessed at the local institutions and, therefore, were treated with trastuzumab, only 98 patients (67.6%) were classified as HER2-positive by the central re-evaluation of HER2 status. Therefore, 47 patients (32.4%) were treated with trastuzumab despite being HER2-negative (Fig. 2).

Selected patient and tumour characteristics for the entire study population and by HER2 status, as defined by the central assessment, are shown in Table II. A total of 69 patients (47.6%) had stage IV disease at the time of diagnosis (de novo MBC), while 52.4% had R-MBC. Most women were of postmenopausal status at the time of trastuzumab initiation (72.4%), while the majority of the patients with R-MBC had received adjuvant chemotherapy (85.5%). The median age at the initiation of trastuzumab-based therapy was 56 years (range, 29–86 years).

In total, 132 patients (91.0%) received trastuzumab as a first-line treatment, while 9 patients (6.2%) were treated with trastuzumab as a second-line treatment. The remaining patients received trastuzumab as third- to seventh-line therapy. In 130 patients (89.7%), trastuzumab was administered with chemotherapy, whereas 13 patients received hormonal therapy at the time of trastuzumab initiation, and 2 patients received trastuzumab as monotherapy. In addition, 6 patients (4.1%) had been previously treated with a trastuzumab regimen in the adjuvant and/or neoadjuvant setting.

The distribution of the markers of interest based on the normalised expression of mRNA-encoding genes is presented in Fig. S1. The distribution of all examined markers by HER2 status (based on central assessment) is presented in Table SII, while Table SIII presents the distribution of markers by ER/PgR status among HER2-positive patients. Table SIV shows the distribution of markers by disease presentation status for the entire study population. HER2-positive patients presented with higher HBEGF, TGFB1 and THRA mRNA expression compared with patients with HER2-negative tumours (P=0.026, P<0.001 and P<0.001, respectively; Table SII), while IGFBP4 mRNA expression was higher in HER2-positive patients with positive ER/PgR status compared with HER2-positive patients with ER/PgR-negative tumours (P=0.004; Table SII). No significant differences were observed in the distribution of the markers of interest between patients with de novo MBC and R-MBC (Table SIV).

IGFBP4 was positively strongly correlated with RARA (rho=0.45; P<0.001), TGFB1 (rho=0.60; P<0.001) and THRA (rho=0.45; P<0.001). In addition, RARA was strongly and positively correlated with THRA (rho=0.52; P<0.001). A strong correlation was also detected between TGFB1 and THRA (rho=0.51; P<0.001) and between BTC and EREG (rho=0.47; P<0.001; Fig. 3).

Patients carrying tumours with high AREG mRNA expression (using the median value as a cut-off point) were of younger age at the time of trastuzumab initiation and were more frequently premenopausal compared with those with low AREG mRNA expression (P=0.002 and P=0.002, respectively). Younger age was also associated with high NRG1 mRNA expression (P=0.030), while lower histological grade was associated with high IGFBP4 mRNA expression (P=0.008). Patients with high mRNA expression levels of RARA, as compared with those with low expression (using the median value as a cut-off point), more frequently exhibited bone metastasis (P=0.012). Higher TIL density was observed in tumours with high TGFA mRNA expression (using the median value as a cut-off point), while PTEN loss was associated with low mRNA expression levels of TGFA and THRA, using the median values as cut-off points (P=0.043, P=0.045 and P=0.003, respectively; Table SV).

The median follow-up time for HER2-positive and HER2-negative patients was 120.3 and 114.2 months, respectively. During this time, 108 patients (74.5%) died (72 HER2-positive, 73.5%; 36 HER2-negative, 76.6%), while 113 (77.9%) experienced disease progression (76 HER2-positive, 77.6%; 37 HER2-negative, 78.7%). The median TTP was 15.1 months (95% CI, 12.6-19.6) and 11.6 months (95% CI, 7.1-17.6) for HER2-positive and HER2-negative patients, respectively. The median survival was 48.5 months (95% CI, 39.2-59.8) for HER2-positive patients and 38.1 months (95% CI, 25.8-49.1) for HER2-negative patients, while no significant differences were observed between HER2-positive and HER2-negative patients in terms of TTP or survival (P=0.33 and P=0.26, respectively).

High NRG1 mRNA expression (using the median value as a cut-off) and high BTC mRNA expression (using the upper quartile as a cut-off) was associated with an increased risk of progression in the HER2-positive population (Table III; Fig. 4). In the HER2-negative population, high EREG mRNA expression (using the median value as a cut-off) was univariately associated with a decreased risk of progression (hazard ratio = 0.45; 95% CI, 0.23-0.90; P=0.025). However, this was not retained upon the adjustment for clinicopathological parameters (P=0.12).

A significant interaction was observed between the disease presentation status and EGF mRNA expression (using the median value as a cut-off) for TTP (interaction P=0.037). High EGF mRNA expression was associated with a decreased risk of progression among patients with de novo MBC (Table III), while the hazard ratio was of the opposite direction in the subgroup of R-MBC women, even though significance was not reached (P=0.43). After adjustment for clinicopathological parameters, a non-significant trend towards improved TTP was observed for patients with de novo MBC carrying tumours with high EGF mRNA expression compared with those with low expression.

In terms of survival, a significant interaction was identified between disease presentation status and EGF expression (using the lower quartile as a cut-off; interaction P=0.045). In the subgroup of patients with de novo MBC, high EGF expression was associated with a decreased risk of death (Table III). No significance was reached among patients with R-MBC (P=0.82).