Ras/Raf-1/MAPK Pathway Mediates Response to Tamoxifen but not Chemotherapy in Breast Cancer Patients

The Ras/Raf-1/mitogen-activated protein kinase (MAPK) pathway controls multiple cellular processes, including proliferation, differentiation, senescence, and apoptosis (1). Activation follows membrane tyrosine kinase receptors (epidermal growth factor receptor, human epidermal growth factor receptor 2-4) binding appropriate ligands, dimerizing, and undergoing autophosphorylation (2, 3). This activates Ras, which translocates to the plasma membrane and promotes Raf-1 phosphorylation and activation (4–13). Raf-1 phosphorylates and activates MEK (MAPK kinase), which activates MAPK, a serine/threonine kinase responsible for phosphorylating and activating substrates, including c-fos and c-myc, which regulate proliferation (14, 15). MAPK also phosphorylates and activates the estrogen receptor α (16–19).

Expression and activation of the Ras/Raf-1/MAPK pathway plays an important role in the development and progression of cancer, and may additionally influence response to treatments targeted against estrogen receptor α and proliferation. Chemotherapy and endocrine therapy are thought to target proliferating cells, and although it remains unclear if highly proliferating tumors are more sensitive to treatment, we hypothesized that activation of this pathway would be an indicator of responsiveness to endocrine and cytotoxic therapies.

The majority of chemotherapeutic agents are thought to function most effectively against proliferating tumor cells in growth phase. Recently the National Epirubicin Adjuvant Trial (NEAT) and the BR9601 trial confirmed that the anthracycline, epirubicin, plus cyclophosphamide, methotrexate, and 5 fluorouracil (CMF), is superior to CMF alone as adjuvant treatment for patients with early breast cancer (20). Because these agents are more effective against rapidly cycling cells (21), it is feasible that tumors with low proliferative indices may be less sensitive to these therapeutic agents. Proliferative indices alone, however, are poor predictors of chemotherapeutic response (22).

Several studies suggest that the Ras/Raf-1/MAPK pathway is linked to response to anthracycline treatment. In vitro Raf-1 (23, 24) and MAPK (25) expression are associated with increased proliferation and anthracycline resistance. Conversely, increased MAPK activation has been linked to enhanced apoptosis and anthracycline sensitivity (26). The role of the Ras pathway in determining chemosensitivity therefore requires further investigation.

Activation of Ras/Raf-1/MAPK is also linked to tamoxifen resistance through phosphorylation of estrogen receptor α. The classic “genomic” action of estrogen receptor α requires ligand binding, which induces phosphorylation, dissociation from heat shock proteins, conformational changes, homodimerization, and nuclear translocation. Nuclear estrogen receptor α binds to ERE sequences in the promoter region of estrogen-regulated genes (27) and recruits coactivators, corepressors, and transcription machinery (16, 28). Estrogen receptor α can, however, be activated in a ligand-independent manner via signaling pathways. The Ras/Raf-1/MAPK pathway phosphorylates Serine 118, within the AF-1 domain of the receptor (16, 18, 19), resulting in its activation and transcription of estrogen-regulated genes and cell proliferation. Consequently, the Ras pathway is thought to increase estrogen receptor α sensitivity to low concentrations of estrogen, resulting in tamoxifen resistance (29–31). Tamoxifen-resistant cells have been shown to have increased levels of activated MAPK, phosphorylated estrogen receptor α, and transcription of estrogen-regulated genes (32, 33).

This study investigated whether expression and activation of the key components of the Ras/Raf-1/MAPK pathway influenced clinical outcome, to test the hypothesis that activation of the pathway drives resistance to tamoxifen and chemotherapy in clinical breast cancer. In order to achieve this, two different patient groups were used, a tamoxifen-treated cohort and a chemotherapy-treated cohort. The study population included estrogen receptor–positive patients who received tamoxifen treatment alone and in combination with chemotherapy, and patients from the Scottish BR9601 trial who received either CMF or sequential epirubicin followed by CMF.

Patient cohorts. Two patient cohorts were studied, after obtaining ethical approval for each study separately. The first comprised 402 patients with estrogen receptor α–positive tumors who were treated at Glasgow Royal Infirmary from 1980 to 1999 [Steroid Resistant Tumour Bank (STB)]. These patients received adjuvant tamoxifen for a median of 5 y (range, 0.6-18 y) and follow-up data were available for a median of 6.45 y (range, 0.64-18.42 y). In addition, 99 (24.8%) of these patients received adjuvant chemotherapy and 110 (27.5%) received adjuvant radiotherapy. Estrogen receptor α status was defined as previously described (34). Clinical/pathologic characteristics for these patients are shown in Table 1. In this study there were 74 breast cancer–specific deaths and 100 breast cancer relapses, 78 of which occurred during tamoxifen treatment.

An additional 318 patients were studied, from the BR9601 adjuvant chemotherapy trial designed to test the possible benefit of four cycles of epirubicin followed by four cycles of CMF over eight cycles of CMF chemotherapy in women with early breast cancer. There was a median follow up of 4.95 y (range, 0.27-8.52 y), with 84 breast cancer–related deaths and 111 breast cancer recurrences (both local and distant). Clinical/pathologic characteristics are again shown in Table 1.

Antibodies. Ras protein expression was investigated using three isoform specific antibodies: H-Ras (IgG₁ Ab, F235; Santa Cruz); K-Ras (Sigma), and N-Ras (IgG1 Ab, F155; Santa Cruz). Raf-1 protein expression was measured using a Raf-1 antibody (IgG1 Ab, E-10; Santa Cruz) and two phospho-specific antibodies recognizing active and inactive Raf-1: phospho-Raf(ser338) (Upstate) and phospho-Raf(ser259) (Cell Signalling Technology), respectively. MAPK expression was investigated using a p44/42MAPK antibody (CST) and a phospho-specific p44/42 MAPK (Thr202/Tyr204) antibody (CST). All Ras antibodies were used at a concentration of 20 μg/mL; the Raf-1 antibody was used at 5 μg/mL, the phospho-Raf-1 antibodies at 4 μg/mL, and both MAPK antibodies at 0.5 μg/mL in antibody diluent (DAKO) for immunohistochemistry. All antibodies were used to investigate protein expression in the STB study but only pRaf(ser259), pRaf(ser338), MAPK, and pMAPK antibodies were used in the BR9601 study. The specificity of all antibodies was confirmed by Western blotting.

Western blotting. Proteins from unstimulated and 10 nmol/L heregulin-stimulated MCF-7 and MDA-MB-453 cells were resolved by 10% SDS-PAGE at 40 mA for 1 h and transferred to polyvinylidene fluoride membrane overnight at 10V. The membrane was treated with 5% bovine serum albumin in TBS-Tween for 1 h and incubated with primary antibody [H-Ras, N-Ras, Raf-1, pRaf(ser338) = 0.4μg/mL, K-Ras, pRaf(ser259), MAPK, pMAPK = 0.2μg/mL] overnight at 4°C. Membranes were incubated in appropriate secondary antibody for 1 h, antimouse IgG (CST; 1:10000) for H-Ras, K-Ras, N-Ras, and Raf-1, and antirabbit IgG (CST; 1:5000) for pRaf(ser259), pRaf(ser338), MAPK, and pMAPK, and visualized using chemiluminescence (Western blotting detection reagent; Amersham Biosciences).

TMA construction. Tissue microarrays were constructed for 402 estrogen receptor α–positive STB tumors and the 318 BR9601 breast tumors (34). Tissue microarray construction allows rapid tumor processing under standardized conditions and has been extensively validated in breast cancer. Up to 100 to 200 (0.6 diameter) individual tumor cores were placed into a single recipient block enabling simultaneous multiple tumor analysis. A consultant breast pathologist was responsible for marking tumor areas on H&E-stained tumor slides prior to coring. To account for tumor heterogeneity, 3 × 0.6 mm cores were removed from the marked areas in each tumor block and transferred to recipient paraffin blocks to form the tissue microarray.

Immunohistochemistry. Immunohistochemistry for H-Ras, N-Ras, and Raf-1 was done as previously described (35). For pRaf(ser338) and K-Ras, antigen retrieval was done by heating under pressure in TE buffer (1 mmol/L EDTA, 5 mmol/L Tris, pH 8.0) for 5 min in a microwave. For pRaf(ser259), p44/42 MAPK, and phospho-p44/42 MAPK, slides were incubated in 10 mmol/L citrate buffer at 96°C for 20 min. Endogenous peroxide was blocked by incubation in 0.3% hydrogen peroxide (H₂O₂), except for K-Ras, in which 3% H₂O₂ was used. Blocking was done using 1.5% normal horse serum (Vector Laboratories; H-Ras, N-Ras, Raf-1, p44/42 MAPK, and phospho-p44/42 MAPK) or Casein solution [Vector Laboratories; K-Ras, pRaf(ser259) and pRaf(ser338)]. Antibody incubations were done overnight at 4°C, with the exception of phospho-p44/42 MAPK (6 h at room temperature). In each run a negative isotype matched and a positive control, using breast tumor tissue known to express the protein of interest, were included. Signal was visualized using Envision (DAKO) and 3,3′-diaminobenzidine (Vector Laboratories). For the K-Ras antibody, the Super Sensitive Non-Biotin HRP Detection System (BioGenex) was used.

Histoscore method. Two observers (LM and TK) trained by a pathologist independently scored tumor cores, as selected by a pathologist, using a weighted histoscore method (35, 36). The intensity of cytoplasmic and nuclear staining was categorized as 0 for negative, 1 for weak, 2 for moderate, or 3 for strong, and the percentage of tumor cells within each category was estimated. The histoscore was calculated using the following formula: Histoscore = 0× % negative tumor cells + 1× % weakly stained tumor cells + 2× % moderately stained tumor cells + 3×% tumor cells stained strongly. The histoscore ranged from a minimum of 0 to a maximum of 300. Agreement between the two observers was monitored. Cases with discordant results between observers were reevaluated. Agreement between observers was excellent with Interclass correlation coefficients scores between 0.74 and 0.97.

Statistical analysis. Statistical analysis was done using the SPSS statistical package (version 9.0 for Windows). Correlations between proteins were calculated using the Spearman rank test. Pearsons χ² test was used to correlate protein expression with known prognostic factors. Kaplan-Meier life-table analysis and Cox's multiple regression (including known prognostic factors tumor size, grade, and nodal status) were done to estimate differences in breast cancer related survival, in terms of disease-free survival (DFS) and overall survival (OS), using breast cancer recurrences and breast cancer–specific deaths as the respective end points. To establish the relative risk of a patient relapsing or dying as a result of either high or low levels of a particular protein in their breast tumor, hazard ratio analysis was calculated by Cox's multiple regression using only the protein of interest as a variable. For survival analysis and χ² tests, patients were split into two groups, those that expressed high levels of protein and those that expressed low levels. For all proteins analyzed, high levels were defined as immunohistochemistry scores equal to or above the upper quartile value, whereas low levels were defined as immunohistochemistry scores less than the upper quartile value. A value of P < 0.05 was deemed statistically significant.

Patient cohort and treatment. Of 402 STB patients, 303 individuals received tamoxifen alone, whereas the remaining 99 received both tamoxifen and chemotherapy. Survival analysis was done on the entire cohort but also on the subgroup of patients who received only tamoxifen, thus addressing the potential confounding effects of both endocrine and chemotherapy treatments. Patients from the BR9601 trial were randomly allocated to receive either CMF alone or epirubicin followed by CMF. Three hundred and eighty-four patients were randomized in BR9601 and tissue samples were retrieved from 318 cases (84%). One hundred and fifty-five (49%) patients received treatment with epirubicin followed by CMF whereas the remaining 163 (51%) patients received only CMF. Additionally, 165 of the BR9601 patients received tamoxifen. A survival analysis of those patients whose samples contributed to this substudy confirmed the statistical advantage of Epi-CMF over CMF observed in the main BR9601 and NEAT studies (ref. 20; data not shown).

Protein expression. H-Ras, K-Ras, N-Ras and Raf-1 expression was investigated in the STB cohort, whereas the inactivated and activated form of Raf-1, pRaf(ser259) and pRaf(ser338) respectively, and MAPK and pMAPK, were analyzed in both the STB and BR9601 cohort of patients. With the exception of pRaf(ser259), which was localized primarily to the cytoplasm, all proteins were expressed in both the cytoplasm and nuclei of tumor cells (Fig. 1, Supplementary Table S1). Despite the high frequency of patients expressing both cytoplasmic and nuclear H-Ras, K-Ras, and N-Ras, there was no significant relationship between the expression levels of Ras in the two locations. A strong positive correlation, however, was evident between the cytoplasmic and nuclear localization of Raf-1, pRaf(ser338), MAPK, and pMAPK.

Activation of the Ras/Raf-1/MAPK pathway in breast tumors. Tumors from the STB and NEAT cohorts with elevated activated Raf, pRaf(ser338), also expressed increased levels of cytoplasmic and nuclear pMAPK (Spearman's rank test; Table 2). In the STB study, overexpression of Ras isoforms was associated with increased activated Raf expression; N-Ras was most markedly correlated with pRaf (P < 0.0005; R² = 0.274, Table 2).

Protein expression and association with known prognostic markers. In the STB cases nuclear pRaf(ser338) expression was positively correlated with node positivity (P = 0.009), and elevated expression of cytoplasmic pRaf(ser338) was associated with increased tumor grade (P = 0.001). Cytoplasmic MAPK expression was positively associated with tumor grade (P = 0.025), size (P = 0.002), and nodal status (P < 0.0005). No correlations were observed between nuclear MAPK and known prognostic markers.

Conversely, in the BR9601 chemotherapy-treated patients no significant correlations were observed between pRaf(ser338) expression and node positivity, tumor size, or grade. Increased nuclear MAPK and cytoplasmic pMAPK expression was related to lower (grade 1 or 2) grade of tumor (P = 0.01 and P = 0.0017, respectively).

STB survival analysis. Raf-1 activation was associated with shortened DFS in the 402 STB patients. High cytoplasmic pRaf(ser338) expression in tumors was associated with a reduced time to recurrence (P = 0.002; Fig. 2A). Patients with tumors expressing increased levels of nuclear pRaf(ser338) also exhibited reduced DFS (P = 0.006). Hazard ratios were 1.84 (95% confidence interval (95% CI), 1.24-2.75; P = 0.0026) and 1.78 (95% CI, 1.17-2.71, P = 0.007), respectively (Fig. 2B). No association was observed between Ras, pRaf(ser259), MAPK, or pMAPK and DFS in this series.

OS for the 402 patients treated in this series was not associated with increased expression of the Ras isoforms, pRaf(ser259), MAPK, or pMAPK. Patients whose tumors expressed increased levels of cytoplasmic or nuclear pRaf(ser338) exhibited a reduced OS following tamoxifen treatment (P = 0.0229 and P = 0.0006, respectively; Fig. 2C and D). Hazard ratios were 1.74 (95% CI, 1.07-2.81; P = 0.0247) and 2.29 (95% CI, 1.41-3.74; P = 0.0009), respectively.

BR9601 survival analysis. Survival analysis revealed no significant association between tumor expression of pRaf(ser259), pRaf(ser338), MAPK, or pMAPK and risk of breast cancer recurrence or death in BR9601 patients receiving chemotherapy (Epi/CMF or CMF) either alone or with tamoxifen.

Estrogen receptor–positive tamoxifen-only–treated cases (303 STB cases) survival analysis. In the 303 STB cases treated only with tamoxifen, patients with tumors overexpressing cytoplasmic (P = 0.0023; 10.2 years versus 13.3 years) or nuclear pRaf(ser338) (P = 0.002; 10.3 years versus 12.8 years) had worse DFS (Fig. 3A and B). The relative risks for relapse associated with increased expression of cytoplasmic or nuclear pRaf(ser338) were 2.02 (95% CI, 1.27-3.21; P = 0.0028) and 2.08 (95% CI, 1.29-3.33; P = 0.0025), respectively. Increased cytoplasmic or nuclear pMAPK expression was associated with a shorter DFS in patients treated with tamoxifen alone (P = 0.01; 8.5 years versus 12.9 years; Fig. 3C; hazard ratio, 2.04 and P = 0.04; 11.4 years versus 12.5 years; hazard ratio, 1.61, respectively). Multivariate Cox-regression analysis revealed that only nuclear pRaf(ser338) expression was independent of tumor size, grade, or nodal status in influencing relapse (P = 0.03; Supplementary Table S2).

Activation of Raf-1 was also linked to a poor outcome in this subset of patients. Elevated cytoplasmic or nuclear pRaf(ser338) expression was associated with shortened OS time (P = 0.015; 13.6 years versus 15.06 years, and P = 0.0008; 11.8 years versus 15.72 years, respectively; Fig. 4A and B). Increased expression of cytoplasmic or nuclear pRaf(ser338) raised the risk of death by 1.96 (95% CI, 1.13-3.42; P = 0.017) and 2.52 (95% CI, 1.44-4.41; P = 0.0012) times, respectively.

Increased expression of MAPK and pMAPK were also associated with a significant reduction in OS time in patients treated only with tamoxifen. Patients whose tumors expressed high levels of cytoplasmic and nuclear MAPK were more likely to die sooner than those with low levels (P = 0.033; 13.2 years versus 15.6 years, and P = 0.039; 13.11 years versus 15.60 years, respectively; Fig. 4C and D). The relative risk for these patients was 1.84 (95% CI, 1.04-3.26) and 1.78 (95% CI, 1.02-31.2) for cytoplasmic and nuclear MAPK, respectively. Likewise, patients with increased tumor levels of nuclear pMAPK exhibited a shortened OS time (P = 0.0336; 13.50 years versus 16.10 years; Fig. 4E). Increased expression of nuclear pMAPK increased the risk of death in patients treated with tamoxifen by 1.83 (95% CI, 1.04-3.24) times. Multivariate analysis excluded pRaf(ser338), MAPK, or pMAPK as being independent of known predictive factors (tumor size, grade, and nodal status) for OS.

Estrogen receptor–positive tamoxifen and chemotherapy–treated cases (STB and BR9601). For the 264 estrogen receptor–positive patients in the combined STB and BR9601 populations who received both tamoxifen and chemotherapy there was no significant association between tumor expression of pRaf(ser259), pRaf(ser338), MAPK, or pMAPK and shortened DFS or OS time (data not shown).

This study shows that the Ras/Raf-1/MAPK pathway is activated in clinical breast tumors and suggests that activation is associated with poor outcome, particularly in patients treated with tamoxifen alone. Analysis of 402 cases treated with tamoxifen suggested that poor outcome was associated with the expression and activation of the Ras cascade, in particular activation of Raf-1 (Fig. 2). In contrast, activation of this pathway was not associated with outcome in a chemotherapy and tamoxifen–treated population from the BR9601 study. This contrasting result from two populations of breast cancer patients, analyzed within the same laboratory, but with differing treatment regimens, led us to hypothesize that activation of Raf-1 may be associated with poor outcome in estrogen receptor α–positive patients treated with adjuvant tamoxifen. We therefore did two exploratory analyses, one on the 303 estrogen receptor α–positive cases from the STB cohort which received only tamoxifen and on the 264 estrogen receptor α–positive cases (99 from the STB and 165 from the BR9601 study) who received chemotherapy and tamoxifen.

Survival analysis showed that increased expression of cytoplasmic and nuclear pRaf(ser338) was associated with increased risk of relapse and death in the 303 tamoxifen-only–treated patients, and on multivariate analysis nuclear pRaf(ser338) expression was independent of nodal status, tumor size, and grade (P = 0.031). In addition, elevated levels of cytoplasmic and nuclear pMAPK were associated with an increased risk of recurrence in this patient cohort. Conversely, in the 264 estrogen receptor α–positive patients treated with both chemotherapy and tamoxifen, no association with phosphorylated Raf-1 or MAPK was observed with patient outcome measures.

These results suggest that increased Raf-1/MAPK phosphorylation or activation is associated with early relapse on adjuvant tamoxifen and that nuclear pRaf(ser338) is a candidate for identifying estrogen receptor α–positive patients at risk of relapse if treated with tamoxifen alone. We cannot at present rule out the possibility that pRaf(ser338) functions as a prognostic marker, as opposed to predictive marker, because all cases analyzed received adjuvant therapy, but it seems to function at least additionally as a predictive factor for improved benefit from chemotherapy. In the 99 STB patients who received both adjuvant tamoxifen and chemotherapy, it seemed that the addition of chemotherapy increased the time to relapse in those patients expressing high levels of pRaf(ser338). Interestingly, pRaf(ser338) was not an independent predictor of tamoxifen resistance when the 303 tamoxifen-only and 99 tamoxifen and chemotherapy–treated patients were combined. This suggests that chemotherapy partially overrides the negative effects of increased expression of nuclear pRaf(ser338). Increased tumor levels of pRaf(Ser338) are perhaps indicative of decreased benefits from tamoxifen but enhanced response to chemotherapy. This supports previous findings that increased expression of Raf-1 in cell lines makes them more responsive to chemotherapeutic agents (37). To confirm this hypothesis, ideally an analysis of an untreated patient cohort would be undertaken to address any potential prognostic role and to determine if pRaf(ser338) is also a predictive marker in the context of benefit associated with tamoxifen therapy.

Raf-1 is a serine-threonine kinase that plays a role in cell proliferation, differentiation, and apoptosis, and is prominent in controlling tumor angiogenesis and metastasis (38, 39). We show that increased Raf-1 activity in tumors was related to poorly differentiated tumors (high grade) and tumor spread (node positivity). A recent study showed that targeting Raf-1 inhibits tumor growth and that this represents an important therapeutic strategy (40).

The ligand-independent phosphorylation of estrogen receptor α at serine 118 by MAPK is believed to be a major contributor to the development of tamoxifen resistance, and this relationship between the Ras/Raf-1/MAPK pathway and estrogen receptor α has been well documented. It is hypothesized that phosphorylation of estrogen receptor α contributes to tamoxifen resistance by promoting tumor growth in the presence of low levels of estrogen (31, 41). In the current study, however, nuclear pRaf(ser338) seems to be dominant over MAPK, which implies that although MAPK-driven phosphorylation of estrogen receptor α may be important, it is not the only contributing factor in the development of tamoxifen resistance. It also suggests that the mechanism by which nuclear pRaf(ser338) influences tamoxifen resistance is independent of MAPK.

In summary, this study shows that expression and activation of the Ras/Raf-1/MAPK pathway in breast tumors is associated with increased risk of relapse and death with tamoxifen treatment but not when chemotherapy is also given to estrogen receptor α–positive cases. These results suggest that activated Raf-1 is a potential predictive marker for identifying patients who are least likely to benefit from tamoxifen and for whom additional therapy may be required.

No potential conflicts of interest were disclosed.

Sadly during the preparation of this manuscript our colleague and friend, Professor Timothy Cooke, died suddenly. We acknowledge his support through the development and execution of this study.