The Ret Receptor Tyrosine Kinase Pathway Functionally Interacts with the ERα Pathway in Breast Cancer

Many members of the receptor tyrosine kinase (RTK) superfamily have essential developmental roles and have been implicated in cancer. The single Ret gene encodes a unique RTK characterized by four repeats of cadherin-like motifs and a juxtamembrane cysteine-rich domain. Alternative splicing of the Ret mRNA results in isoforms that differ in the last coding exon; the two major isoforms Ret9 and Ret51 having 9 and 51 specific amino acids, respectively. Although these isoforms are commonly coexpressed, they exert specific functions (1). Ret is the receptor for the glial-derived neurotrophic factor (GDNF) family of growth factors, also comprising neurturin, artemin, and persephin. These ligands bind Ret in conjunction with glycosylphosphatidylinositol (GPI)-anchored coreceptors of the GDNF receptor α (GFRα) family. GFRα1-to-4 form homodimers that are recruited by a preferential ligand into a high-affinity complex, which in turn interacts with and activates Ret homodimers. Alternatively, Ret is activated in trans through shed (GFRα1) or soluble (GFRα4) versions of the coreceptors (2), leading to robust signaling and potentiation of biological effects in some model systems (3).

Ret is essential for development of the sympathetic, sensory, and enteric nervous system, where it promotes survival, differentiation, and migration. Furthermore, Ret is expressed in the nephric ducts and nascent buds, where GDNF acts as a morphogen to trigger ureteric bud outgrowth. Ret-deficient mice die shortly after birth from severe defects in enteric neuron and glial cell development, in addition to kidney agenesis. Interestingly, the phenotype of Ret-, GFRα1-, and GDNF-deficient mice are remarkably similar, although not completely overlapping (2, 4), suggesting a major influence of the GDNF/GFRα1 complex on Ret physiologic functions.

Ret is a paradigm of a single gene that causes different types of human neuroendocrine cancers when targeted by different mutations. Various germ line, gain-of-function mutations trigger three dominantly inherited cancer syndromes affecting neuroendocrine tissues: multiple endocrine neoplasia type 2A (MEN2A), type 2B, and familial medullary thyroid carcinoma. Somatic Ret mutations are also frequently observed in sporadic medullary thyroid carcinoma. Furthermore, sporadic or radiation-induced rearrangement of Ret results in papillary thyroid carcinomas. These malignancies are each characterized by the production of constitutively activated versions of the Ret RTK, resulting therefore in enforced activation of various signaling pathways (5–7).

Conversely, little is known regarding Ret expression, activity, and a potential role in progression of tumors from non-neuroendocrine origin. In this report, we have examined the expression of Ret RTK pathway components in breast tumor cell lines and primary human tumors. Ret is expressed in many breast tumor cell lines, in particular, those characterized by estrogen receptor α (ERα) expression and/or ErbB2 overexpression. GDNF treatment of ERα-positive cell lines leads to increased oncogenicity and Ret-dependent, anchorage-independent proliferation, with a functional interaction with the ERα pathway shown. Finally, we provide evidence that Ret signaling components are expressed in primary breast tumors. In summary, our data support the potential of Ret as a novel therapeutic target in human breast tumors.

Cell lines. BT474, HCC1419, HCC1937, HCC1954, HCC2218, MCF7, MDA-MB134, MDA-MB231, MDA-MB361, MDA-MB453, SKBR3, T47D, and ZR75.1 breast tumor cell lines and the Bing packaging cell line were obtained from the American Type Culture Collection. MDA-MB175 breast tumor cell line was obtained from Dr. M. Bentires-Alj (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland).

Reagents and human tissue sources. Recombinant human GDNF and GFRα1 were obtained from PeproTech EC and R&D Systems, and reconstituted at 100 μg/mL in water and PBS supplemented with 1% bovine serum albumin, respectively. E2 (Sigma-Aldrich) was prepared at 1 mmol/L in ethanol.

The Ret inhibitor NVP-BBT594 was obtained from Dr. P. Manley (Novartis Institute for Biomedical Research, Basel, Switzerland) and was prepared at 10 mmol/L in DMSO.

Human breast cancer tissues were obtained from the University Hospital, Lausanne, Switzerland.

Anchorage-independent proliferation. Plates were coated with the cationic polymer polyHema (12 mg/mL in 95% ethanol) to prevent cell adhesion. Single cell suspensions (10⁶) were immediately stimulated with the ligand or treated with the Ret inhibitor before ligand stimulation for 96 h. SiRNA-transfected MCF7 cells (0.5–0.6 × 10⁶) were seeded on polyHema-coated plates 48 h posttransfection and immediately stimulated with GDNF for an additional 96 h. Viable cells were counted with a ViCell Coulter Counter (Beckman Coulter, Inc.). For BrdUrd incorporation experiments, proliferating cells in suspension were treated with 10 μmol/L BrdUrd for 3 h before flow cytometry analysis using the FITC BrdUrd Flow kit (Becton Dickinson).

Soft agar. Proliferating cells (10⁵) were stimulated in suspension for 10 to 20 min, then made up to 0.4% agar in culture medium and seeded onto a 0.8% agar layer. Embedded cells were refed with medium plus ligand every week for 4 wk. Colonies were stained with nitrobluetetrazolium salt, and colonies larger than 0.5 mm in diameter were counted with an Artek colony counter (Dynatech Laboratories, Inc.).

Steroid deprivation. MCF7 and T47D cells was steroid deprived using phenol red–free DMEM supplemented with 10% charcoal-stripped FCS (Hyclone) for 4 (suspension) or 6 d (monolayer), before stimulation with 1 nmol/L 17-β estradiol (E2). Treatments were initiated either immediately (in suspension) or 2 d postseeding (monolayer), and cells were treated every other day for 6 d for the proliferation assays. For mRNA expression analyses, steroid-deprived cells were treated for 24 h plus 48 h with 1 nmol/L E2 or vehicle. For protein expression analyses, steroid-deprived cells were treated for either 8 or 24 h with 1 nmol/L E2 or vehicle.

Statistical analysis. Fisher's Exact test, One-, or Two-Way ANOVA statistical analysis using Tukey test for pairwise comparisons were performed as appropriate using SigmaStat 3.1 (Systat Software GmbH). A P value of <0.05 was considered statistically significant.

The Ret RTK is expressed in breast tumor cell lines. To investigate whether Ret is expressed in breast cancer cell lines, real-time TaqMan primers specific for the two major isoforms Ret9 and Ret51 were used to screen a panel of 15 lines. MCF7, MDA-MB134, MDA-MB175, MDA-MB361, HCC1419, and HCC2218 expressed the highest levels of Ret mRNA, whereas moderate to low levels were detected in BT474, MDA-MB453, SKBR3, T47D, and ZR-75.1 (Fig. 1A). No expression was found in HCC1937, HCC1954, MDA-MB231, and MDA-MB468. With the exception of the BT474, HCC1419, and MDA-MB361, the two major isoforms were similarly expressed. These data show Ret mRNA expression in the majority (11 of 15) of the tumor cell lines tested. Consequently, Ret9 protein was analyzed in this panel using an antibody specific for the short isoform (Fig. 1B); Ret9-MEN2A overexpressing NIH-3T3 was used as a positive control. The 150 and 170 kDa differentially glycosylated forms were detected in the majority of cell lines positive for Ret mRNA (Fig. 1B,, arrows). A similar expression pattern was observed for Ret51 (data not shown). With the exception of BT474, there was a concordance between Ret mRNA and protein expression. Antibody specificity was confirmed using siRNAs to reduce Ret9 and Ret51 protein expression in MCF7 and T47D cells. Cells were transfected with control siRNAs (siLuc and siLacZ, respectively) or with specific siRNAs targeting all Ret isoforms (siRet1 and siRet2; data not shown; Fig. 1C). Importantly, bands attributed to Ret9 and Ret51 were dramatically decreased after Ret-specific siRNA transfection, compared with control transfections.

Further analysis of the molecular features of the screened tumor cells indicated that most of the lines with detectable Ret mRNA and protein expression also exhibited ERα expression and/or ErbB2 amplification (P = 0.009 versus ERα expression and ErbB2 amplification; Fig. 1A). However, in this relatively small panel of cell lines, neither ERα nor ErbB2 is a significant independent factor that determines Ret expression. Finally, Ret expression was undetectable in the three lines negative for ERα and not ErbB2 amplified (HCC1937, MDA-MB468, and MDA-MB231). Taken together, these data show for the first time Ret protein expression in a significant portion of breast tumor cell lines.

GDNF triggers signaling pathway activation in breast tumor cell lines in a Ret-dependent manner. In the present study, we selected the ERα-positive, ErbB2-independent MCF7 and T47D lines for functional analyses of the Ret signaling pathway. The prototype coreceptor GFRα1 was detected in MFC7, but not in T47D cells (Fig. 1D), indicating the need for concomitant treatment of T47D cells with soluble recombinant forms of GDNF and GFRα1. Accordingly, MCF7 and T47D cells were stimulated with increasing concentrations of human GDNF alone (MCF7) or in the presence of human GFRα1 (T47D). Importantly, we provide evidence for increased tyrosine phosphorylation of Ret9 upon ligand stimulation (Fig. 2A and B); in particular, as little as 10 ng/mL GDNF was sufficient to trigger maximal Ret phosphorylation in T47D cells (Fig. 2B). Furthermore, in MCF7 cells, we observed a slight increase in Ret9 phosphorylation specifically on residue Tyr1062 (Fig. 2A), a residue shown to be essential for Ret signaling in vivo (8, 9). Ret protein level was not affected by short-term GDNF stimulation in both lines (Figs. 1C and 2A and B). The increase in Ret phosphorylation upon GDNF stimulation is relatively limited compared with the potent downstream effects on Erk phosphorylation (Fig. 2A and B). It is possible that after GDNF treatment, other tyrosine residues, for which there are currently no specific antibodies, become more strongly phosphorylated than Tyr1062.

Ret mediates its biological functions through the activation of various intracellular signaling pathways (9–12). Time-dependent effects of increasing concentrations of GDNF (1, 10, and 100 ng/mL) were, therefore, analyzed alone or in combination with increasing concentrations of GFRα1 (10, 100, and 1,000 ng/mL), in serum-deprived MCF7 and T47D cells, respectively. GDNF stimulation elicited a strong increase in Erk phosphorylation at concentrations as low as 1 and 10 ng/mL in MCF7 (Fig. 2A and C) and T47D (Fig. 2B and D), respectively. However, the kinetics and the duration of activation differed. Thirty minutes of treatment was required to trigger maximal activation in T47D (Fig. 2D), whereas Erk phosphorylation could be detected 5 minutes after ligand addition and was maintained for several hours in MCF7 (data not shown). Furthermore, there was a transient enhancement of c-Jun-NH₂-kinase (JNK) phosphorylation in both lines. In contrast to most Ret-driven model systems (5–7), GDNF treatment of these cell lines resulted in little or no increase in Akt phosphorylation at the time points analyzed (data not shown). Similar effects of GDNF on these signaling events were also observed in complete medium (Fig. 3A–C). In summary, these data define 10 ng/mL GDNF alone or in combination with 100 ng/mL GFRα1 as the optimal concentration promoting activation of downstream signaling in breast cancer cell lines.

It has been previously described that GDNF can activate signaling pathways in the absence of Ret expression or engagement (2, 5). Therefore, we addressed Ret dependency of GDNF-induced signaling in breast cancer cells using a Ret tyrosine kinase inhibitor (NVP-BBT594; Supplementary Fig. S1A) and siRNA/shRNA approaches. In the constitutively active Ret-dependent thyroid TT model, NVP-BBT594 efficiently lowered receptor phosphorylation and decreased cellular proliferation with an IC₅₀ of 5 to 10 nmol/L (Supplementary Fig. S1B and C). Importantly, pretreatment of MCF7 and T47D cells with low inhibitor concentrations completely prevented GDNF-induced Erk phosphorylation (Fig. 3A). Furthermore, siRNA-mediated down-regulation of Ret in MCF7 cells dramatically reduced GDNF-induced Erk phosphorylation, whereas the control β-galactosidase siRNA had no effect (Fig. 3B). ShRNA-mediated stable down-regulation of Ret in T47D cells (Supplementary Fig. S2A) also resulted in dramatic abrogation of Erk activation by GDNF (Fig. 3C). Altogether, the signaling data clearly show that Ret is functional in breast cancer cell lines and mediates GDNF-elicited molecular effects.

Ret enhances anchorage-independent proliferation. Treatment of MCF7 and T47D breast cancer cells with GDNF had only marginal mitogenic activity when cells were cultivated on plastic (data not shown). Furthermore, treatment with the Ret inhibitor at concentrations that prevented GDNF-induced signaling did not affect proliferation (Supplementary Fig. S3). These results suggest that Ret does not significantly modulate proliferation or survival of breast cancer cells maintained as a monolayer. Accordingly, we tested whether GDNF can induce anchorage-independent growth of MCF7 and T47D cells using soft agar methodology (Fig. 4A). In the absence of GDNF, few colonies formed when cells were seeded in a solid phase agar layer. Strikingly, weekly GDNF stimulation (100 ng/mL) significantly enhanced colony formation by 8 and 4.7 times in MCF7 and T47D cells, respectively. Therefore, these results indicate that GDNF triggers increased survival and/or proliferation of single cells grown in soft agar.

To further investigate the contribution of Ret to the proliferation and/or survival of breast cancer cells grown in three dimensions, we cultivated the cells in suspension on plates coated with polyHema, a cationic polymer that prevents adhesion to the substratum. When seeded on polyHema-coated plates in complete medium, MCF7 and T47D cells tightly aggregated within a few hours, resisted anoikis, and proliferated at a slow rate (data not shown). In this setting, GDNF increased Erk phosphorylation in single cells (freshly seeded) and cell aggregates (24 hours postseeding), an effect specifically prevented in cells pretreated with the Ret inhibitor NVP-BBT594 (Fig. 3D).

Consistent with the increased growth in soft agar and ligand-dependent signaling in suspension, GDNF treatment of MCF7 and T47D cells seeded in polyHema-coated plates significantly enhanced anchorage-independent proliferation; increasing viable cell number after 96 hours incubation by 30% and 70%, respectively (Fig. 4B). More specifically, GDNF triggered anchorage-independent mitogenic activity, as measured by BrdUrd incorporation (Fig. 4D), with no evidence of changes in basal anoikis levels (data not shown). Furthermore, the proliferative effect of GDNF was prevented by NVP-BBT594 pretreatment in a concentration-dependent manner (Fig. 4B). Similarly, transient down-regulation of Ret by specific siRNAs before seeding MCF7 cells in polyHema-coated plates, or stable down-regulation of Ret by shRNAs in T47D cells efficiently blocked the proliferation enhancement by GDNF (Fig. 4C). Taken together, these results show that Ret signaling stimulates anchorage-independent proliferation of breast cancer cells.

Ret potentiates estrogen-driven proliferation. As Ret expression is associated with ERα expression in breast tumor cell lines, a possible interaction between ER and Ret signaling pathways was addressed. Initially, mRNA expression of Ret signaling components in response to E2 stimulation was evaluated by semiquantitative PCR. Estrogen-deprived MCF7 and T47D cells were left untreated or stimulated with E2 for 3 days. Strikingly, in both lines, Ret and GFRα1 mRNA levels were dramatically increased upon E2 treatment (Fig. 5A,, top). This long-term E2 stimulation resulted only in moderate up-regulation of the well-established, estrogen early-responsive genes TFF1, cyclin D1, and c-myc (13–16) in a cell line–specific manner, with increased expression of TFF1 and of cyclin D1 and c-myc observed in T47D and MCF7 cells, respectively (Fig. 5A). Interestingly, in response to short-term E2 treatment, Ret and GFRα1 mRNA were rapidly induced with kinetics similar to those of the early-responsive genes TFF1 and c-myc (Supplementary Fig. 4), suggesting that Ret and GFRα1 might be direct target genes for estrogen signaling. Importantly, short-term E2 stimulation of steroid-deprived cells resulted in a major increase in Ret protein levels, which were maintained at 24 hours in both cell lines (Fig. 5A , bottom).

Next, Ret signaling in estrogen-deprived and stimulated settings was compared. Twenty four hours after E2 addition, GDNF treatment induced Erk phosphorylation, which was inhibited by NVP-BBT594 indicating Ret dependency (Fig. 5B). However, in deprived cells, Erk was only moderately phosphorylated in response to GDNF, presumably because of limiting Ret amounts in the absence of estrogens. Thus, estrogens induce sustained expression of Ret and GFRα1, which contributes to enhanced GDNF signaling.

The above data strongly suggest a functional interaction between the Ret and ER pathways in breast tumor cell lines. As estrogens stimulate MCF7 (Supplementary Fig. S5) and T47D (Fig. 5C) proliferation when cells are maintained on plastic and in suspension, a possible contribution of Ret signaling to E2-driven proliferation was assessed. Strikingly, GDNF treatment significantly enhanced E2-driven proliferation of T47D cells maintained on plastic (Fig. 5C,, left) and in suspension (Fig. 5C,, right). This effect is mediated by Ret, as it is abolished after stable Ret down-regulation (Fig. 5D). This potentiation was not observed in MCF7 cells cultivated on plastic (Supplementary Fig. S5). In summary, these data show an interaction between Ret and ER signaling in T47D and MCF7 breast cancer cells, which in the former case is associated with Ret-enhanced E2-driven proliferation. The reason for the differential response of the two cell lines is currently unknown. One possibility is that they differ in their dependence on endocrine- versus Ret-driven proliferation, a possibility that remains to be explored.

Ret is expressed in primary human breast tumors. To support the functional contribution of the Ret signaling pathway in breast tumor biology, a panel of 10 primary breast tumor biopsies and MCF7 xenografts were screened for expression of Ret signaling components by real-time PCR (Fig. 6). For normalization, the level of Ret mRNA in T47D cells was set as one. Ret9 and Ret51 were expressed in two independent extracts of MCF7 xenografts. Importantly, all the breast tumors had detectable expression of Ret mRNA, with similar levels of Ret9 and Ret51. Four of 10 primary tumors exhibited high Ret expression levels (≥4-fold the level in T47D cells; Fig. 6). It is noteworthy that these four are ERα positive, suggestive of a possible association between Ret levels and ERα status. Furthermore, GDNF mRNA was expressed at variable levels in all the tumors and MCF7 xenografts. Furthermore, GFRα1, which was only expressed in 2 of 15 of the original breast tumor cell lines (MCF7 and MDA-MB361; data not shown), was detected in 9 of 10 primary tumors and in both xenografts. Coexpression of GDNF and GFRα1 with the Ret receptor in some tumors is suggestive of a functional Ret signaling pathway in a subset of human breast cancers.

Ret is the driving oncogene in various neoplasms of the thyroid, where specific mutations lead to defined tumor types. A potential role for Ret in non-neuroendocrine tumor biology has not been widely addressed. In contrast to the situation in thyroid tumors, Ret activating mutations have not been reported in high-throughput mutation profiling of human breast cancers (17). In the present study, we have reported Ret expression in primary breast tumors and Ret signaling activity in breast tumor cell lines.

Differential signaling by cis-acting, GPI-anchored and trans-acting, soluble GFRα1 forms has been documented (2), with long-lasting biological effects triggered by trans stimulation in neuronal cells. In MCF7 and T47D breast cancer models, low GDNF concentrations stimulated Ret phosphorylation and signaling, resulting in Erk and JNK activation. Interestingly, the duration of pathway activation differed notably. In contrast, in this study, cis-activation in MCF7 triggered more sustained signaling than trans-activation in T47D, suggestive of tissue type– or tumor-specific differences in the regulation of Ret-modulated signaling pathways.

The Erk pathway has been shown to mediate mutant Ret effects on the proliferation and migration of thyroid tumor cells (18). Moreover, the importance of this pathway in papillary thyroid tumor development is shown by the mutually exclusive mutation pattern in the Ret/Ras/B-Raf axis (19). Similarly, JNK activation has been implicated in Ret-dependent mitogenic and motile activities (20). In breast tumor cells, GDNF had no obvious mitogenic effect in standard culture conditions on plastic; however, it significantly enhanced the proliferation of cells maintained in suspension in a Ret-dependent manner. Furthermore, GDNF significantly increased anchorage-independent growth in soft agar, an activity that has not previously been described for Ret ligands. These results indicate that the Ret pathway modulates the oncogenicity of breast tumor cells and might contribute to their invasive potential (21).

GDNF and another Ret ligand artemin have been implicated in pancreatic tumor cell invasion in vitro (22, 23). In combination with its preferred coreceptor GFRα3, artemin signaled to Erk in MCF7 (not in T47D cells; data not shown) but had no effect on anchorage-independent proliferation (Supplementary Fig. S6). These results indicate some level of specificity in Ret ligand activities in these breast tumor models. Importantly, we observed coexpression of Ret and GFRα1 mRNA in a number of primary tumors. This is consistent with a previous report indicating overexpression of GFRα1 mRNA specifically in breast tumors (24). Furthermore, our detection of GDNF mRNA in most breast tumor biopsies indicates that all components of the major axis for Ret activation are coexpressed, suggesting that this pathway might be functionally relevant for breast tumor biology.

Importantly, in breast tumor cell lines, Ret mRNA was detected in ERα-positive lines. In support of our observations, it has recently been reported that Ret mRNA is significantly up-regulated in human ERα-positive primary breast tumors (25). This prompted us to analyze the interaction between the Ret and ER pathways at the molecular and functional level. In this respect, we showed that Ret mRNA and protein expression is positively regulated by estrogen signaling in T47D and MCF7 cells, as previously suggested by transcriptional profiling studies (16, 25). Interestingly, a recent whole-genome cartography of ERα binding sites (ERE) in MCF7 cells identified Ret as a putative ERα direct target gene (26), consistent with our observation that E2 stimulation rapidly up-regulated Ret mRNA. Furthermore, although GFRα1 was not described in the whole-genome ERE cartography (26), we found its expression to be estrogen responsive. These results indicate coordinated transcriptional regulation for Ret signaling components, thus strengthening the potential relevance of Ret signaling in endocrine-dependent breast tumor biology. With this in mind, it is of interest to note that in the presence of E2, Ret activation significantly enhanced E2-driven proliferation of T47D cells.

In our limited analysis of primary breast tumors, higher Ret levels were detected in ERα-positive tumors, consistent with a recent profiling study (25). Expression profiling from the Amsterdam National Cancer Institute publicly available database of 295 primary breast tumors also indicated strong association between higher Ret expression (or higher GFRα1 expression) with ERα-positive status (courtesy of Dr. Martin Buess, Hospital University Basel, Switzerland). Therapeutics targeting ERα or aromatase have made a major effect in the prevention of endocrine-driven breast cancer progression; however, resistance eventually occurs in most patients and is often associated with up-regulation and activation of growth factor signaling pathways (27). Importantly, our results showing a functional interaction between the ERα and Ret pathways may, therefore, be important to consider in the light of progression of tumors and/or resistance to hormonal therapy. These possibilities are worthy of further investigation.

In summary, we show that Ret is expressed in primary breast tumors and is functional in a panel of breast tumor cell lines, in particular, cell lines characterized for their dependence on endocrine signaling. There are a number of small molecule inhibitors that have shown potent Ret inactivation in vitro and in preclinical models; some are currently being tested in Ret-dependent thyroid tumor patients (28). Our data support the relevance of directly addressing the effect of Ret signaling inhibitors in breast cancer. Specifically, the functional interaction with ERα signaling in vitro strengthens the potential of Ret pathway inhibition in a combination approach with endocrine therapy.

No potential conflicts of interest were disclosed.

Grant support: Swiss Cancer League (OCS-01445-12-2003; C. Brisken, M. Fiche, and N.E. Hynes) and the Novartis Research Foundation (N.E. Hynes).

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.

We thank Drs. Carlos Garcia-Echeverria and Giorgio Caravatti [Novartis Institutes for BioMedical Research (NIBR), Basel, CH] for reviewing the manuscript; Dr. Clare Isacke (Breakthrough Breast Cancer Research Centre, Institute of Cancer Research, London, UK) and Dr. Martin Buess (Hospital University, Basel, Switzerland) for helpful discussions; Dr. James Fagin (Division of Endocrinology and Metabolism, University of Cincinnati, Ohio) for supplying the TT cells; Dr. Johannes Roesel (NIBR, Basel, CH) for supplying the NPM-ALK Ba/F3 cells; Dr. James Griffin (Dane Farber Cancer Institute, Boston) for supplying the RET/PTC3, Tel-IGF1R, and ErbB2(V659E) Ba/F3 cells; Dr. Paul Manley (NIBR, Basel, CH) for supplying the Ret inhibitor NVP-BBT594; and Dr. Terence O'Reilly (NIBR, Basel, CH) for expert advice on the statistical analyses.