Gemcitabine Synergizes with Immune Checkpoint Inhibitors and Overcomes Resistance in a Preclinical Model and Mesothelioma Patients

This article is featured in Highlights of This Issue, p. 6105

T cells can recognize and control transformed cells (1); however, protective immunity against cancer often is insufficient due to a plethora of immune escape mechanisms, including immune checkpoints such as programmed death receptor 1 (PD-1, CD279; ref. 2). PD-1 is expressed on the surface of activated T cells (3) and interaction with its ligands PD-L1 (CD274, B7-H1) or PD-L2 (CD273, B7-DC; ref. 4) results in T-cell unresponsiveness (5). PD-L1 can be expressed by T and B cells, myeloid cells, endothelial cells, and by various tumors (6), whereas the expression of PD-L2 is limited to dendritic cells (DC), macrophages, and mast cells (4).

Blockade of the interaction between PD-1 and its ligands has shown remarkable clinical responses in a proportion of patients with melanoma or non–small cell lung cancer (NSCLC; refs. 7, 8). Also in mesothelioma, treatment with immune checkpoint inhibitors (ICI) is promising with responses in about 20% (9) and disease control rate in 50% of patients (presented at ESMO 2017, Zalckman G., abstract LBA58_PR). Indeed, the efficacy of ICI in mesothelioma has been shown in different studies (9, 10). It is still a matter of debate whether the expression of PD-L1 by the tumor cells might be a useful predictive marker for survival or for clinical responsiveness to treatment with anti-PD-L1 (11–13). These discordant data may be explained by the difficulty to faithfully stain tumor samples for PD-L1, as well as the availability of multiple staining protocols, which hampered conclusive interpretations regarding the prognostic relevance of PD-L1 expression in many cancer types including mesothelioma (14, 15).

In this study, we show that the expression of PD-L1 by mesothelioma cells has prognostic power in a large cohort of mesothelioma patients. Furthermore, we demonstrate feasibility and efficacy of combining gemcitabine with ICI in a preclinical model for mesothelioma. We chose to combine gemcitabine with ICI because of its common use in mesothelioma and immunostimulating effects (16–18). Finally, we tested this combination in two patients, who did not respond to pembrolizumab or gemcitabine as monotherapy, and observed clinical response in both cases.

Large sections from 145 patients with mesothelioma were collected between 1999 and 2009. Patients received either a combination treatment with cisplatin/gemcitabine (with cisplatin on day 1 and gemcitabine on day 1 and 8, every three weeks) or cisplatin/pemetrexed (both on day 1, every 3 weeks). Postchemotherapy tumor samples were collected after three cycles of treatment. In total, we collected 251 samples of which 147 were of the epithelioid, 14 of the biphasic and 90 of the sarcomatoid subtype. The 251 samples were collected as follows: 117 were surgical biopsies, 132 were surgical specimens, and 2 were needle-core biopsies. From those 251 samples, 118 were collected before and 133 after chemotherapy; 160 of these samples were matched samples from 80 patients, taken before as well as after chemotherapy. From this cohort of 145 patients, matching samples of 80 patients before and after chemotherapy were available. Clinical data are summarized in Supplementary Table S1. All patients gave informed, written consent and the study was conducted in accordance with the declaration of Helsinki. The study was approved by the Institutional Ethical Review Board of the University Hospital Zurich under reference number StV 29-2009 and EK-ZH 2012-0094.

Formalin-fixed paraffin-embedded (FFPE) blocks of the abovementioned samples were retrieved from the archives of the Institute of Surgical Pathology, University Hospital of Zurich (Zurich, Switzerland). All cases were reviewed by two pathologists (B. Vrugt and V. Tischler) on full sections of a representative tumor block and classified following the definitions of the WHO classification for each subtype of mesothelioma. In case of biphasic mesothelioma, areas were identified that clearly separated the epitheloid and sarcomatoid growth patterns and 2 cores were taken from each area.

Tissue microarrays (TMA) were constructed with a custom-made, semiautomatic tissue array (Beecher Instruments) as described previously (19, 20). The TMAs were built using at least two tissue punches with cores of 0.6 mm diameter for each patient using a Ventana ES instrument (Ventana Medical Systems; ref. 21). TMA blocks were sectioned and stained with hematoxylin and eosin for morphologic assessment. Additional cores of control tissues, including 8 tonsils, 8 normal lung tissues, 8 normal pleura and 8 adenocarcinomas of the lung, were included in duplicate. Technical factors leading to loss of cores during TMAs processing (i.e., sectioning and staining) resulted in a reduction of the patient numbers available for subsequent statistical analysis.

In addition, we analyzed 20 matching whole tumor sections from 10 patients, taken before and after chemotherapy. These samples are included in the TMA described above. Those patients were treated with cisplatin [75 mg/m², every three weeks (Q3W)] plus pemetrexed (500 mg/m² Q3W) as part of a clinical trial (22).

The following human mesothelioma cell lines were used in this study: ZL55 (epitheloid), SPC111 (sarcomatoid), and MSTO-211H (biphasic). ZL55 and SPC111 were generated in our laboratory (23) and MSTO-211H (24) was obtained from ATCC. Cells were cultured in humidified incubator at 37°C and 5% CO₂ in either DMEM (for ZL55 and SPC11) or RPMI (for MSTO-211H) supplemented with 10% heat-inactivated FBS, 2 mmol/L l-glutamine and antibiotics (all from Gibco). Cells were tested negative for Mycoplasma ssp. using the VenorGeM Mycoplasma detection kit (Sigma).

Gemcitabine (Eli Lilly and Company), cisplatin (Sigma Aldrich), and pemetrexed (Eli Lilly and Company) were obtained from the local pharmacy and dissolved in PBS. Final concentrations in cell culture were as follows: gemcitabine, 0.1, 0.2, 0.5, 1, and 10 μmol/L; cisplatin, 0.01, 0.1, and 1 μmol/L; pemetrexed, 0.1, 1, and 10 μmol/L. The drug was added at the start of culture and once more after 24 hours. Media were used as negative control (25). Analysis was performed 48 hours after the start of cultures.

C57BL/6 mice were obtained from Harlan Laboratories (Envigo) and kept under specific pathogen-free conditions at the Laboratory Animal Services Center (LASC) of the University of Zurich (Zurich, Switzerland). Six- to 8-week-old female mice were used for all experiments. Experiments were performed in accordance with the Swiss federal and cantonal regulations on animal protection and were approved by the Cantonal Veterinary Office Zurich.

RN5 cells are crocidolite-induced mesothelioma cells of C57BL/6 origin (26, 27). RN5 cells were tested negative for Mycoplasma ssp. using the VenorGeM Mycoplasma detection kit (Sigma). C57BL/6 mice were injected subcutaneously with 10⁶ RN5 cells in 100 μL PBS. Tumor growth was measured every 3–4 days in two dimensions (length and width) with a caliper. When tumors reached a size of 40–50 mm² (approximately 40 days after tumor cell injection), mice were randomized into different treatment groups and treatment was started. The death event was defined as tumor size reaching the legally acceptable limit of 225 mm².

Gemcitabine (Eli Lilly and Company) was dissolved in PBS and given intraperitoneally (i.p.) once per week at a dose of 120 mg/g body weight (28). Anti–CTLA-4 (clone UC10-4F10-11) and anti-PD-1 (clone RMPI-14) were purified from culture supernatant using protein G Sepharose 4 Fast Flow (GE Healthcare) columns according to the manufacturer's protocol. Antibodies were administered intraperitoneally once per week at 250 μg/100 μL PBS (29). Dexamethasone (Sigma-Aldrich) was administered orally once per week at 0.3 mg/kg body weight (30). All treatments were given on the same day. This experiment was repeated two times with at least 5 mice/group.

Cultured cells were harvested by 0.5% Trypsin-EDTA treatment and stained with the following fluorophore-labeled mAbs and appropriate isotype controls: PD-L1 (clone 29E.2A3), PD-L2 (clone 24F.10C12), HLA-ABC (clone G46- 2.6), HLA-DR (clone L243). Dead cells were stained using Zombie Violet Fixable Viability Kit. Anti-HLA-ABC and anti-HLA-DR were obtained from BD Biosciences and all other reagents from BioLegend. Cells were incubated with appropriately diluted antibodies in PBS for 25 minutes at 4°C. Subsequently, cells were washed and fixed for 5 minutes with 2% paraformaldehyde in PBS. Samples were measured on a CyAn ADP9 machine (Beckman Coulter) and analyzed using FlowJo v9.8.5 software (Tree Star).

For IHC, cell lines were pelleted after in vitro treatment, and formalin-fixed, paraffin-embedded, sectioned, and subjected to IHC for PD-L1 using clone E1L3N as described below under the section “IHC.”

The use of gemcitabine plus pembrolizumab was justified by the fact that both drugs are approved as monotherapy. The combination treatment was used off-label with written informed consent of the patients according to the declaration of Helsinki and was approved by the University Hospital Zurich (Zurich, Switzerland).

For IHC, 4-μm thick paraffin sections were cut from the TMA block and mounted on silane-coated glass slides. The sections were stained using an automated IHC platform (Ventana Benchmark). The following antibodies were used: anti-PD-L1 (E1L3N, Cell Signaling Technology), anti-FOXP3 (236A/E7, Abcam), anti-CD3 (clone LN10, Leica Biosystems), anti-Perforin (clone 5B10, DBS). Anti-PD-L1 staining with E1L3N antibody was performed by pretreatment of samples in H₂O₂ for 30 minutes and then samples were incubated with the antibody diluted as 1:1,000 at room temperature for 30 minutes and heat-induced epitope retrieval HIER2 solution with a Bond Polymer Refine Detection kit (Leica Biosystems). Samples with <50 evaluable cells were excluded. Biopsies were considered PD-L1⁺ when >1% of tumor cells were stained positive with anti-PD-L1. Comparison of PD-L1 staining using either the clone E1L3N from Cell Signaling Technology or the clone SP142 from Ventana showed a correspondence in 100% of cases either negative or positive (Supplementary Fig. S1). Each core was evaluated independently by two pathologists (D. Soldini and V. Tischler). We used archived FFPE samples. Although the use of archival FFPE samples to determine PD-L1 expression in mesothelioma has not yet been validated, it was shown that matching fresh and FFPE lung cancer samples gave similar results regarding PD-L1 expression determined by IHC (31).

Resected tumors were processed for FFPE and 4-μm thick sections were stained for IHC by Sophistolab AG using a Leica BondMax instrument and Refine HRP-Kits (Leica DS9800, Leica Microsystems Newcastle) according to the manufacturer's guidelines. The following antibodies were used: PD-L1 (Lifespan LS-B9795), FoxP3 (Novus Biologicals, NBP1-18319), CD3 (RMAB005, clone SP7, DBiosystem). Hematoxylin counterstaining was performed according to standard protocol. Whole slides were scanned with a Zeiss Mirax Midi slide scanner (20× objective, NA0.8) equipped with a 3-CCD color camera (Hitachi HV-F22) and analyzed using Pannoramic viewer 1.15.4 (3DHISTECH).

The Kaplan–Meier method was used to estimate tumor-specific overall survival (OS). Tumor-specific OS in humans was determined from the date of histologic diagnosis of mesothelioma to the date of death from the same cancer up to 5 years of follow-up. Patients, who were alive at the time of last follow-up, were censored for OS analysis. Continuous data were compared between groups using the Mann–Whitney U test. For comparison of experimental groups, two-tailed Student t test with Welch correction was performed. Statistical significance P < 0.05. Statistical analyses were performed using IBM SPSS Statistics 20.0 (SPSS Inc)

For mice, tumor-specific OS was determined from the date of randomization and the death event was defined as tumor size reaching the legally acceptable limit of 225 mm². For survival analysis, the Log-rank test (Mantel Cox test) was used. For comparison between each experimental groups ANOVA with Bonferroni correction was performed. Statistical analyses were performed using PRISM (GraphPad Prism version 7 for Mac, GraphPad Software).

We investigated the expression of PD-L1 in a cohort of 145 patients with mesothelioma (Supplementary Table S1). Tumor samples were available from 100 cases before chemotherapy and 122 cases after chemotherapy; 80 of these cases were paired samples taken from the same patient before and after chemotherapy. We detected expression of PD-L1 in 14% of samples before chemotherapy (14 of 100) and in 18% of cases after chemotherapy (22 of 122; Supplementary Table S2); expression of PD-L1 was membranous and homogenous, and the intensity of expression varied from low (+) to high (+++; Fig. 1A and B). We found that expression of PD-L1 on tumor cells was a negative prognostic marker (P = 0.04) for survival independent of both the histologic subtype and other clinical features (Fig. 1C).

Using large sections, we detected CD3⁺ T cells throughout the tumor tissue, but none of them expressed perforin. Instead, most of the CD3⁺ T cells expressed FOXP3⁺ and were preferentially localized in PD-L1⁺ tumor areas (Supplementary Fig. S2).

To investigate whether chemotherapy with cisplatin/pemetrexed or cisplatin/gemcitabine has an impact on the expression of PD-L1 by mesothelioma cells, we compared the expression of PD-L1 in paired samples taken before and after chemotherapy from 80 patients. Although we found that the PD-L1 expression differed between matched samples, these changes seemed random (Supplementary Table S2), suggesting that the chemotherapies investigated here had no straightforward impact on PD-L1 expression. From the cohort of 80 paired samples taken before and after chemotherapy, we further analyzed 20 large sections (i.e., paired samples from 10 patients) for expression of PD-L1 on tumor cells and confirmed the data described above (not shown).

To directly address whether chemotherapeutic drugs influence the expression of PD-L1 on mesothelioma cells, we incubated two human PD-L1⁺ mesothelioma cell lines (MSTO and SPC11) and one human PD-L1⁻ mesothelioma cell line (ZL55) with gemcitabine or cisplatin/pemetrexed. We did not observe chemotherapy-induced changes with respect to PD-L1 expression by FACS in any of the cell lines (Fig. 2). To further validate our data, we made cell blocks of the previously mentioned conditions and perform IHC for PD-L1 (Supplementary Fig. S3). There was a complete correlation between FACS and IHC for MSTO (positive) and ZTL55 (negative) cell lines (either untreated, treated with gemcitabine or treated with cisplatin/pemetrexed). The SPC11 cell line (positive by FACS) was weakly positive only in one of the three samples (untreated, treated with gemcitabine, or treated with cisplatin/pemetrexed). This is consistent with reports showing variation among different PD-L1 antibodies (32) and that PD-L1 analysis by FACS and IHC is not fully overlapping (33).

Together, these data suggest that neither cisplatin/pemetrexed nor cisplatin/gemcitabine induce direct changes in PD-L1 expression by mesothelioma cells. Rather, we think that the random differences observed between paired samples may be explained by intratumoral heterogeneity regarding PD-L1 expression or by indirect effects of chemotherapy. For example, some chemotherapies including cisplatin and doxorubicin induce immunogenic cell death and result in local stimulation of tumor-specific immunity (34) and this will produce IFNγ at the tumor site. PD-L1 expression over time could be then the result of local IFNγ production (35). Furthermore, our data suggest that standard-of-care chemotherapy does not per se preclude subsequent treatment with PD-1–blocking antibodies.

To investigate whether gemcitabine synergizes with ICI, we used a mouse mesothelioma model, subcutaneous RN5 tumors in syngeneic C57BL/6 mice. This model shares histopathologic features with the aggressive human mesothelioma subtype (sarcomatoid). We found that treatment with gemcitabine impeded tumor progression and promoted infiltration by CD3⁺ T cells, whereas treatment with ICI had no significant impact on tumor progression or survival. The combination of gemcitabine and ICI, however, significantly prolonged survival and resulted in tumor rejection in a proportion of the mice (Figs. 3A and B), suggesting a synergistic effect of ICI plus gemcitabine.

Dexamethasone is an anti-inflammatory and immunosuppressive glucocorticoid that is commonly used together with chemotherapy as an antiemetic drug. Specifically, dexamethasone is often given with gemcitabine to prevent side effects including chemotherapy-induced nausea and vomiting, fever, and chills. In contrast to brief treatment schedules, long-term use of glucocorticoids is immunosuppressive (24). We observed a therapeutic synergy when ICI was combined with gemcitabine, suggesting that activation of immune defense supports the clinical response to gemcitabine. Therefore, we wondered whether long-term treatment with dexamethasone would negatively influence this therapeutic synergy and treated mice with established RN5 tumors with gemcitabine plus ICI or gemcitabine plus ICI plus dexamethasone. Mice treated with gemcitabine alone or vehicle served as controls. Again, we observed that gemcitabine plus ICI resulted in significantly better survival compared with gemcitabine alone (Fig. 3B). However, addition of dexamethasone to the treatment with gemcitabine plus ICI significantly reduced responses in treated mice (Fig. 3B). Next, we investigated the immune infiltrate in mouse mesothelioma by IHC for CD3 and FoxP3 at the endpoint. We observed increased fibrotic and necrotic areas after treatment with gemcitabine plus ICI, as well as increased number of infiltrating CD3⁺ T cells compared with all other groups (Fig. 4). Furthermore, none of the CD3⁺ cells expressed FoxP3 (data not shown). In contrast, tumors of mice treated with gemcitabine plus ICI plus dexamethasone did not show necrotic areas and very few infiltrating CD3⁺ T cells (Fig. 4), suggesting that corticosteroids negatively influence the clinical response to gemcitabine plus ICI. We think that this may apply to any other therapy that involves immune stimulation including standard therapies (30, 34), suggesting that the use of steroids in the context of cancer therapies should be avoided if possible. It may very well be of general applicability that clinical responses to ICI are stronger in absence of steroids. The finding that concomitant use of steroids was associated with poor outcomes of patients with NSCLC treated with ICI (36) supports this idea.

Currently, the standard of treatment of mesothelioma is chemotherapy. However, blockade of the interaction between PD-1 and its ligands has shown promising clinical responses in a proportion of patients in clinical trials (9). We present here two patients with mesothelioma who were refractory to various therapies but responded to a combination of anti-PD-1 and gemcitabine.

The first patient was 66 years old and diagnosed with epithelioid mesothelioma three years ago (in 2014). The patient underwent multimodality approaches including chemotherapy and tumor resection. Because of relapse of disease, the patient was treated with pembrolizumab. Upon progression, the patient was treated with carboplatin/pemetrexed. Then he was switched to carboplatin/gemcitabine and consecutively to gemcitabine monotherapy. Because of further progression, the patient received a combination of gemcitabine (1,000 mg/m² weekly) plus pembrolizumab (200 mg every 3 weeks) without corticosteroids. Before starting this treatment, the patient complained about dysphagia and weight loss. However, two months after starting this combination treatment, the patient could eat normally and a radiologic and metabolic response to treatment was detected (Fig. 5A and B). He is still under therapy, 20 weeks after the start of treatment.

The second case was a 57-year-old patient diagnosed with epithelioid mesothelioma 4 years ago (in 2013). The patient underwent multimodality approaches including chemotherapy and tumor resection (Fig. 5C). Because of recurrence of disease in 2016, the patient was treated with pembrolizumab (anti-PD-1), which resulted in progression of disease (Fig. 5D, time point i), and subsequently with carboplatin/pemetrexed with further progression of disease (Fig. 5D, time point ii). Then one symptomatic lesion in the chest wall was irradiated with a total dose of 36 Gy (6 × 6 Gy). Restaging was performed (Fig. 5D, time point iii) before start of weekly chemotherapy with gemcitabine (without corticosteroids). Within two months of gemcitabine monotherapy, there was evidence of progression of the mediastinal mass and lung metastasis (Fig. 5D, time point iv) and the patient had a reduced performance status with ECOG 3 due to cachexia and dyspnea. Because the patient did not show any limiting side effects on gemcitabine, we started a treatment with pembrolizumab (200 mg every three weeks) and gemcitabine (1,000 mg/m² weekly), in March 2017. After three cycles of this treatment, the patient clinically improved and went back to work using his bicycle as means of transport. Furthermore, radiologic images showed a clinical response of the mediastinal mass and lung metastasis with increased pleural effusion (Fig. 5D, time point v). We observed no side effects besides grade 2 anemia and pleural effusion that was drained after treatment. We analyzed the composition of the pleural effusion and observed that together with few mesothelioma cells there was 80% of lymphocytes (Figs. 6A and B), of these, 90% were CD3⁺ T cells, consisting of 20% CD4⁺ (Fig. 6C) and 75% CD8⁺ cells (Fig. 6D). This suggests that the combination of gemcitabine plus anti-PD-1 mobilized protective immunity resulting in an objective clinical response. Because of his improved general condition, the patient wished to stop the treatment and undertake a journey of 5 weeks. Upon his return, treatment with gemcitabine plus anti-PD-1 was started with a three-weekly schedule for both medications as per the patient's request. After two cycles of this treatment, a progression of disease was detected and the treatment was discontinued.

Combining different therapies for cancer is a promising approach to overcome resistance and improve responses. For example, combination of nivolumab (anti-PD-1) with ipilimumab (anti-CTLA-4) has shown improved clinical responses compared with monotherapy; however, also increased side effects (37). Currently, combinations of different chemotherapies with immunotherapy are under evaluation for different tumor types (NCT02039674, NCT02366143). We investigated whether the combination of gemcitabine with ICI is an effective and feasible treatment for mesothelioma that outperforms monotherapy and may even overcome resistance to monotherapy.

Expression of PD-L1 by tumor cells is often used as inclusion criterion for treatment with antibodies that block the interaction of PD-1 with its ligands (9). Therefore, we first investigated whether gemcitabine influences the expression of PD-L1 by mesothelioma cells in vivo and in vitro. In addition, we found that the expression of PD-L1 is a negative biomarker for survival in mesothelioma, which goes in line with previous studies (14, 38) and with data obtained from other tumor entities (39–42).

We found that treatment with gemcitabine plus ICI significantly prolonged survival of mesothelioma-bearing mice compared with gemcitabine alone, whereas ICI as monotherapy had no impact. Although ICI as monotherapy shows clinical efficacy in patients with mesothelioma (43), this is not the case in our preclinical model. Absence of clinical efficacy of ICI as monotherapy has been described in multiple mouse models (29) but this discrepancy with humans is not understood.

Corticosteroids are commonly used to manage side effects of chemotherapy. Because of their immunosuppressive properties, we evaluated the impact of dexamethasone on the clinical efficacy of gemcitabine plus ICI and observed an abrogation of therapeutic synergy. This suggests that side effects of chemotherapy should be managed by other medications than corticosteroids, whenever possible. Indeed, we have successfully applied this approach to both patients presented here.

Our data show that PD-L1 expression is heterogeneous and prone to immune response–induced changes. In addition, various studies have shown a high degree of intratumoral heterogeneity in PD-L1 expression in different cancers (44, 45). Therefore, we think that using PD-L1 expression as an inclusion criterion for treatments targeting this pathway is not sufficiently justified for patients with mesothelioma. We demonstrated the synergy of gemcitabine with ICI providing a rationale for a clinical trial that can include patients with different tumor types, where gemcitabine represents a standard of care and might be favorable in terms of tolerability and costs compared with combination of several immunomodulating agents. It has been shown that in patients with advanced lung cancer that daily administration of steroids reduces lymphocyte numbers (46) and may be associated with worse outcome (36). Obviously, highly emetogenic treatments must be combined with antiemetic drugs, however, corticosteroids should be avoided and established alternatives used instead. Here, we show in a preclinical model that avoiding the use of dexamethasone improves the efficacy of immunotherapy for mesothelioma, a disease with very limited effective treatment options (47) and an unmet need for new therapies (48).

C. Britschgi is a consultant/advisory board member for Roche, AstraZeneca, and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Tallón de Lara, P. Stolzmann, M. van den Broek, A. Curioni-Fontecedro

Development of methodology: E. Felley-Bosco, R.A. Stahel, V. Tischler, M. van den Broek, A. Curioni-Fontecedro

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Tallón de Lara, V. Cecconi, S. Hiltbrunner, B. Bode-Lesniewska, I. Opitz, P. Stolzmann, C. Britschgi, D. Soldini, M. van den Broek, A. Curioni-Fontecedro

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Tallón de Lara, V. Cecconi, M. Friess, B. Bode-Lesniewska, B. Vrugt, R.A. Stahel, V. Tischler, D. Soldini, M. van den Broek, A. Curioni-Fontecedro

Writing, review, and/or revision of the manuscript: P. Tallón de Lara, S. Hiltbrunner, H. Yagita, B. Bode-Lesniewska, I. Opitz, W. Weder, P. Stolzmann, E. Felley-Bosco, C. Britschgi, D. Soldini, M. van den Broek, A. Curioni-Fontecedro

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Cecconi, H. Yagita, M. Friess, I. Opitz, P. Stolzmann, M. van den Broek, A. Curioni-Fontecedro

Study supervision: M. van den Broek, A. Curioni-Fontecedro

This work was financially supported by the Sophien Stiftung Zurich, the Swiss National Science Foundation, the University of Zurich (University Research Priority Project “Translational Cancer Research”), the Forschungskredit of the University of Zurich and the Science Foundation for Oncology (SFO). We thank André Fitsche, Susanne Dettwiler, and Martina Storz (Institute for Surgical Pathology, University Hospital Zurich, Zurich, Switzerland) for excellent technical assistance with TMA construction and establishment of IHC staining. We thank the personnel of the Laboratory Animal Services Center (LASC) at the University of Zurich for expert animal care.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.