Purpose: Activation or overexpression of HER-2/neu is associated with up-regulation of vascular endothelial growth factor (VEGF) in human breast cancer cells in vitro. Preclinical experiments indicate that increased expression of VEGF may in part mediate the biologically aggressive phenotype of HER-2/neu-overexpressing human breast cancer. It was the purpose of this study to: (a) evaluate the association between HER-2/neu and VEGF expression in a large clinical cohort of primary breast cancer patients; (b) compare the prognostic significance of VEGF isoforms; and (c) analyze the combined effects of HER-2/neu and VEGF on clinical outcome.
Experimental Design: HER-2/neu and VEGF were measured by ELISA in primary breast tumor tissue lysates from 611 unselected patients with a median clinical follow-up of 50 months. At least six VEGF isoforms consisting of 121, 145, 165, 183, 189, or 206 amino acids are generated as a result of alternative splicing. The VEGF121–206 ELISA uses antibodies that bind to VEGF121 and, therefore, detects all of the VEGF isoforms with 121 and more amino acids. The VEGF165–206 ELISA uses antibodies that bind to VEGF165 and, therefore, detects all of the VEGF isoforms with 165 and more amino acids. VEGF121–206 and VEGF165–206 were analyzed both as continuous and categorical variables, using detectable expression as a cutoff for positivity. Cell lines with defined HER-2/neu expression levels were used to establish a cutoff point for HER-2/neu overexpression in breast tumor samples.
Results: Our findings indicate a significant positive association between HER-2/neu and VEGF expression. VEGF121–206 and VEGF165–206 expression was detectable in 88 (77.2%) and 100 (87.7%), respectively, of the 114 patients with HER-2/neu-overexpressing tumors, in contrast to 271 (54.5%) and 353 (71.0%), respectively, of the 497 patients with nonoverexpressing tumors (χ2 test: P < 0.001 for both VEGF121–206 and VEGF165–206). VEGF121–206 and VEGF165–206 demonstrate a comparable prognostic significance for survival in unselected primary breast cancer patients (univariate analysis: VEGF121–206, P = 0.0068; VEGF165–206, P = 0.0046; multivariate analysis: VEGF121–206, P = 0.1475; VEGF165–206, P = 0.1483). When the analyses were performed separately for node-negative and node-positive patients, VEGF121–206 and VEGF165–206 were of prognostic significance for survival only in node-positive patients (univariate analysis: VEGF121–206, P = 0.0003; VEGF165–206, P = 0.0038; multivariate analysis: VEGF121–206, P = 0.0103; VEGF165–206, P = 0.0150). A biological concentration-effect relationship between VEGF expression and survival (VEGF121–206, P = 0.0280; VEGF165–206, P = 0.0097) suggests that VEGF levels, as determined by ELISA, could be of importance as a predictive marker for therapeutic strategies that target VEGF. Combining HER-2/neu and VEGF121–206/VEGF165–206 results in additional prognostic information for survival (VEGF121–206, P = 0.0133; VEGF165–206, P = 0.0092).
Conclusion: The positive association between HER-2/neu and VEGF expression implicates VEGF in the aggressive phenotype exhibited by HER-2/neu overexpression, and supports the use of combination therapies directed against both HER-2/neu and VEGF for treatment of breast cancers that overexpress HER-2/neu.
Vascular endothelial growth factor (VEGF) is a secreted heparin-binding glycoprotein and one of the most potent endothelial cell-specific mitogens. It is known to play a key role in tumor angiogenesis (1 , 2) . Expression of VEGF is subject to a number of control mechanisms, including induction and potentiation in response to hypoxia (3) , numerous cytokines (interleukins 1 and 6), transforming growth factor α and transforming growth factor β (2) , inactivation of tumor suppressor genes such as p53 (4 , 5) , and activation of oncogenes such as ras (6) or src (7 , 8) . The HER-2/neu proto-oncogene encodes a transmembrane tyrosine kinase receptor, is amplified and overexpressed in 20–25% of human breast cancers (9, 10, 11, 12) , and is associated with up-regulation of VEGF in vitro (13, 14, 15) . Preliminary transfection studies indicate that engineered overexpression of HER-2/neu is associated with increased expression of VEGF in human breast cancer cells at both the RNA and protein levels in vitro (14 , 15) . Moreover, exposure of these cells to recombinant heregulin β-1 significantly increases VEGF protein secretion into conditioned medium, whereas exposure to the anti-HER-2/neu antibodies, 4D5, or trastuzumab significantly decreases VEGF specifically in HER-2/neu-overexpressing cells in vitro (13 , 15 , 16) . Taken together, these preclinical data suggest that VEGF is a downstream target of the HER-2/neu signaling pathway; however, clinical data on the interaction between HER-2/neu and VEGF are very limited and inconclusive. Therefore, it was the primary purpose of the current study to explore the association between HER-2/neu and VEGF by measuring the exact protein concentration of HER-2/neu, VEGF121–206, and VEGF165–206 by ELISA in a consecutive series of 611 patients treated for primary breast cancer. A comprehensive comparison of the VEGF expression in HER-2/neu-overexpressing and -nonoverexpressing patients, and subsequent analysis of these results provides a clearer understanding of the relationship between HER-2/neu and VEGF.
Human VEGF mRNA is transcribed from eight exons, and at least six VEGF isoforms consisting of 121, 145, 165, 183, 189, or 206 amino acids are generated as a result of alternative splicing from a single VEGF gene (17, 18, 19, 20, 21) . Stimpfl et al. (22) demonstrated that all of the splice variants except VEGF206 are expressed in breast cancer tissue specimens and cell lines, with variants VEGF121 and VEGF165 being the most common isoforms in breast cancer. VEGF145 has only rarely been detected in breast cancer cells, and its expression levels were similar to VEGF183 and VEGF189, being 10–100-fold lower than those of VEGF121 and VEGF165 (22) . The various VEGF isoforms differ in their binding capacities to tyrosine kinase receptors [VEGFR-1 (flt-1) and VEGFR-2 (KDR/flk-1)], and the neuropilin receptor (18 , 20 , 23, 24, 25, 26) . VEGF121 is freely soluble, whereas VEGF165 can be bound to cell membranes or extracellular matrix, and heparin or proteolytic enzymes can support the activation of VEGF165 by freeing it from its bound state (27) . Despite these biochemical differences, the physiological significance of different isoforms remains unclear. In vivo data suggest that the VEGF121 isoform is more strongly tumorigenic than VEGF165 (28) ; however, this has not yet been demonstrated clinically. Previous studies on the prognostic relevance of VEGF in breast cancer have measured either VEGF121–206 or VEGF165–206 levels, but have not directly compared both isoforms in the same cohort. The epitope for the VEGF121–206 antibody binds to the VEGF amino acid 121, which is present on all six of the isoforms, and as a result, this antibody detects all of the isoforms. The VEGF165–206 antibody detects an exon 7-encoded peptide not present in the 121 and 145 isoforms, and, therefore, can measure only the VEGF165 isoform and those with greater molecular weight. A secondary objective of the current study was to compare the prognostic relevance of VEGF121–206 and VEGF165–206, either alone or in combination with HER-2/neu, for disease-free survival (DFS) and overall survival (OS) among node-positive and -negative breast cancer patients.
An association between HER-2/neu overexpression and VEGF expression in a clinical cohort could confirm the hypothesis based on preclinical evidence that increased VEGF expression may be a result of HER-2/neu overexpression and may contribute to the aggressive phenotype of HER-2/neu-overexpressing breast cancer cells. This would provide a rationale for combined therapeutic approaches targeting both VEGF and HER-2/neu.
MATERIALS AND METHODS
Breast cancer tissue specimens originated from 611 consecutive unselected patients treated for breast cancer at the Department of Obstetrics and Gynecology, University of Munich, Klinikum Grosshadern between 1992 and 1997. The study cohort was originally assembled to evaluate the prognostic significance of urokinase-type plasminogen activator or plasminogen activator inhibitor in primary breast cancer, and has been described and characterized previously, with a median patient follow-up of 26 months (29 , 30) . In the current study, the follow-up has been extended to 50 months for all of the patients, 52 months for node-negative patients, and 49 months for node-positive patients. Follow-up information was available for 603 (98.7%) of the 611 patients. Criteria for exclusion from the study included distant metastasis at the time of diagnosis. Patients underwent either modified radical mastectomy or lumpectomy with complete axillary lymph node dissection, followed by radiation therapy of residual breast tissue. All of the lymph node-positive patients received adjuvant chemotherapy and/or tamoxifen therapy, whereas lymph node-negative patients received adjuvant chemotherapy and/or tamoxifen therapy only if adverse prognostic factors were present, such as large primary tumors (>2 cm), negative hormone receptor status, or elevated levels of either urokinase-type plasminogen activator or plasminogen activator inhibitor. Tumor samples and clinical information were obtained under Institutional Review Board approval.
Tissue Extraction and ELISAs for HER-2/neu and VEGF.
Breast cancer tissue specimens were selected by the pathologist at the time of surgery and stored at −70°C until use. Frozen tissue samples of 100–200 mg were pulverized with a microdismembrator (Braun-Melsungen, Melsungen, Germany) for 30 s at maximum power, suspended in 900 μl Tris-buffered saline (pH 8.5) containing 1% Triton X-100 detergent (Sigma, Munich, Germany), incubated at 4°C for 12 h with gentle shaking, and ultracentrifuged at 100,000 × g for 45 min to remove cell debris, nuclei, and membranes. HER-2/neu protein levels were measured with a commercially available ELISA kit (Oncogene Research Products, Cambridge, MA) and are reported as fmol/mg protein.
The concentrations of VEGF121–206 and VEGF165–206 were retrospectively analyzed in tumor lysates routinely prepared for steroid hormone receptor measurements. Tissue samples of 200–500 mg collected at the same time as samples used to measure HER-2/neu were homogenized with an Ultra-Turrax homogenizer (Jahnke and Kunkel, Staufen, Germany) in 2–3 ml of buffer containing 10 mm Tris, 0.1% monothioglycerol acetate, and 1.5 mm EDTA (pH 7.4). The homogenate was then centrifuged for 30 min at 200,000 × g at 4°C and VEGF measured in the resulting supernatant using quantitative ELISAs. VEGF expression was analyzed as a continuous variable, as well as a categorical variable. For the latter purpose, VEGF content was compared between two patient groups with either detectable VEGF levels or no measurable VEGF expression. The VEGF121–206 ELISA uses a mouse monoclonal antibody A4.6.1 for capture and a biotinylated monoclonal antibody 2E3 for detection. Both antibodies bind to VEGF121; therefore, this assay detects all of the VEGF isoforms with 121 or more amino acids (VEGF121, VEGF145, VEGF165, VEGF183, VEGF189, and VEGF206). The assay also detects recombinant VEGF with amino acids 8–109, making it likely to detect VEGF110, a know product of plasmin digestion of VEGF165 (31) . Using recombinant VEGF165 as a standard, the standard range for the VEGF121–206 ELISA was 42–5200 fmol/ml. Because a minimum of 1:10 dilution of the tumor lysate samples was performed, the lowest assay sensitivity of the VEGF121–206 ELISA was 420 fmol/ml. Values below this detection limit were classified as VEGF-negative. The VEGF165–206 ELISA uses a mouse monoclonal antibody (3.5F8) that detects exon 7-encoded peptide sequences specific for VEGF165 for capture and biotinylated A4.6.1 for detection (32) . This ELISA measures VEGF isoforms of ≥165 amino acids, and the standard range for the VEGF165–206 ELISA was 26–3328 fmol/ml. The lowest assay sensitivity of the VEGF165–206 ELISA was 260 fmol/ml. Values below this de-tection limit were classified as VEGF-negative.
Cell lines with defined HER-2/neu expression levels were used to establish a cutoff point for the quantitative HER-2/neu ELISA in breast tumor samples. We published previously that among human breast cancer cell lines that do not exhibit HER-2/neu gene amplification or overexpression, HER-2/neu expression ranged between 61.5 (95% confidence interval, 45.8–76.3) fmol/mg protein for MDA-MB-231 cells and 297 (95% confidence interval, 268–326) fmol/mg protein in MCF-7 cells (30) . We investigated the association between HER-2/neu overexpression and VEGF expression at multiple cutoff points over a range of 300–500 fmol/mg protein at intervals of 50 fmol/mg protein. HER-2/neu overexpression was positively associated with VEGF expression at each of the selected cutoff points (data not shown). For additional analyses, we chose 400 fmol/mg protein as a representative cutoff point to define HER-2/neu overexpression. Moreover, a HER-2/neu value of 400 fmol/mg protein in the present cohort, with a median follow-up of 50 months, provided separation of patients with regard to DFS and OS [univariate analysis: DFS, P = 0.0001, risk ratio (RR) 1.92 (95% confidence interval, 1.38–2.66); OS, P = 0.03, RR 1.56 (95% confidence interval 1.04–2.32)].
Associations between categorical variables were analyzed by the χ2 statistic. Associations between continuous variables were calculated by Spearman’s ρ correlation. Absolute levels of VEGF were compared by the Mann-Whitney U test between HER-2/neu-overexpressing and -nonoverexpressing patients. When tested as a continuous variable in univariate or multivariate analyses, the VEGF data were log-transformed, due to unequal variance. The prognostic significance of HER-2/neu, VEGF, and other clinical/pathological variables was determined using a univariate Cox model and the log rank statistic. Prognostic independence was analyzed using a multivariate Cox regression model. VEGF165–206 or VEGF121–206, tumor size, and age were included in the model as continuous variables; nuclear grade (1, 2 versus 3, 4), hormone receptor status [estrogen receptor (ER) and/or progesterone receptor (PR) -positive versus ER- and PR-negative], number of involved axillary lymph nodes (0 versus 1–3 versus >3), and HER-2/neu status (<400 versus ≥400 fmol/mg protein) were included as categorical variables. Statistical calculations were performed using SPSS (SPSS, Inc., Chicago, IL) or SAS (SAS, Inc., Cary, NC) software. All of the tests of statistical significance were two-sided.
To investigate the association between HER-2/neu overexpression and VEGF expression, and to compare the clinical relevance of the two most abundantly expressed VEGF isoforms, HER-2/neu, VEGF121–206, and VEGF165–206 expression were studied in a cohort of 611 unselected primary breast cancer patients. Patient demographic data, as well as disease characteristics of the study cohort, are shown in Table 1⇓ . These data demonstrate that the cohort is typical of the general primary breast cancer population, as reported previously (12 , 33 , 34) . VEGF expression was measured by ELISA, using an antibody that detects VEGF isoforms of ≥121 amino acids and with an antibody that recognizes the VEGF isoforms of ≥165 amino acids.
VEGF121–206 expression was detectable among 359 of 611 (58.8%) of the patients, and VEGF121–206 protein levels ranged from 208 to 147,368 fmol/mg. VEGF165–206 expression was detectable among 453 of 611 (74.1%) of the patients, and VEGF165–206 protein levels ranged from 130 to 21,736 fmol/mg. HER-2/neu overexpression (HER-2/neu levels of at least 400 fmol/mg protein) was seen in 114 of 611 (18.7%), and HER-2/neu levels ranged from 0–12,015 fmol/mg protein.
Association between VEGF and HER-2/neu.
VEGF121–206 was detectable in 88 of 114 (77.2%) of the patients with HER-2/neu-overexpressing tumors, in contrast to 271 of 497 (54.5%) of patients with HER-2/neu-nonoverexpressing tumors (P < 0.001), and patients with HER-2/neu-overexpressing tumors had significantly higher VEGF121–206 concentrations as compared with patients with HER-2/neu-nonoverexpressing tumors [median and interquartile range, 1157 (0–2522) versus 442 (0–1326); P < 0.001].
VEGF165–206 expression was detectable in 100 of 114 (87.7%) of the patients with HER-2/neu-overexpressing tumors, as opposed to 353 of 497 (71.0%) of patients with HER-2/neu-nonoverexpressing tumors (P < 0.001), confirming a positive association between HER-2/neu overexpression and VEGF165–206 expression. Again, patients with HER-2/neu- positive tumors had significantly higher VEGF165–206 concentrations as compared with patients with HER-2/neu-nonoverexpressing tumors [median and interquartile range, 523 (276–952) versus 416 (0–806); P = 0.015]. To investigate whether the positive association between HER-2/neu overexpression and VEGF expression was sensitive to the selected cutoff for HER-2/neu overexpression, additional cutoff values between 300 and 500 fmol/mg protein at 50-fmol/mg protein intervals were tested. At each of these additional cutoff values, we observed a comparable significant positive association (P ≤ 0.001) between HER-2/neu overexpression and VEGF121–206 or VEGF165–206 expression (data not shown).
After the observation that the frequency of VEGF expression was significantly higher among patients with HER-2/neu-overexpressing tumors compared with patients with HER-2/neu-nonoverexpressing tumors, we investigated whether absolute levels of HER-2/neu and VEGF121–206 or VEGF165–206 expression were correlated among those patients with tumors that express VEGF and overexpress HER-2/neu. Surprisingly, neither the absolute VEGF121–206 nor VEGF165–206 expression levels were correlated with the absolute HER-2/neu protein levels among patients with HER-2/neu-overexpressing tumors and detectable VEGF expression (VEGF121–206 r = 0.08, P = 0.476; VEGF165–206 r = −0.003, P = 0.975), indicating that absolute VEGF levels are likely subject to numerous additional control mechanisms, including hypoxia (3) , activation of oncogenes (6 , 7 , 8) , multiple cytokines (2) , and loss of tumor suppressor function (4 , 5) .
Associations between VEGF and Other Tumor Variables.
VEGF121–206 expression, analyzed as a categorical variable, was associated with tumor size of >2 cm (P = 0.002), a negative hormone receptor status (ER and PR <10 fmol/mg; P < 0.001), high nuclear grade (G3-G4; P < 0.001), invasive ductal histology (P < 0.001), and positive axillary lymph node status (P = 0.015; Table 2⇓ ). When VEGF121–206 and absolute estrogen or progesterone receptor levels, tumor size, and number of positive axillary lymph nodes were analyzed as continuous variables, VEGF121–206 concentrations positively correlated with tumor size (r = 0.13; P = 0.001) and inversely correlated with estrogen (r = −0.10; P = 0.018) and progesterone receptor levels (r = −0.14; P < 0.001), but did not correlate with the absolute number of positive axillary lymph nodes.
VEGF165–206 expression, analyzed as a categorical variable, was associated with the same clinical/pathological variables as VEGF121–206, such as tumor size of >2 cm (P < 0.001), high nuclear grade (G3-G4; P < 0.001), and invasive ductal histology (P < 0.001), but not with hormone receptor or lymph node status. As a continuous variable, VEGF165–206 correlated positively with tumor size (r = 0.16; P < 0.001), but not with hormone receptor expression levels or the number of positive axillary lymph nodes.
To explore the prognostic significance of VEGF, HER-2/neu, and other clinical/pathological variables, univariate analysis was performed for the entire group of patients with follow up information (n = 603), and for node-positive (n = 304) and node-negative (n = 288) patient subsets. Detectable VEGF121–206 expression was a statistically significant predictor of poor DFS and OS among the entire patient group (n = 603; DFS, P = 0.0459, RR, 1.37 (1.01–1.86); OS, P = 0.0328, RR, 1.49 (1.03–2.15); Table 3⇓ ). Moreover, when VEGF121–206 was analyzed as a continuous variable, absolute VEGF121–206 levels were significantly correlated with DFS and OS [DFS, P = 0.0232, RR, 1.07 (1.01–1.12); OS, P = 0.0068, RR, 1.10 (1.03–1.17)]. Detectable VEGF165–206 expression also demonstrated prognostic significance for OS (P = 0.0495, RR, 1.55 (1.01–2.40)] but not DFS [P = 0.3559, RR, 1.18 (0.83–1.66); Table 3⇓] . When VEGF165–206 was analyzed as a continuous variable, absolute levels of VEGF165–206 were significantly correlated with OS. However, unlike VEGF121–206 there was no correlation between VEGF165–206 levels and DFS [DFS, P = 0.1218, RR, 1.07 (0.98–1.15); OS, P = 0.0046, RR, 1.15 (1.05–1.27)].
Other parameters that demonstrated prognostic significance for DFS and OS in univariate analyses included HER-2/neu status, age, tumor size, number of positive lymph nodes (0 versus 1–3, versus >3), hormone receptor status (ER- and/or PR-positive versus ER- or PR-negative), and nuclear grade (G1, 2 versus G3, 4; Table 3⇓ ).
When univariate analysis was restricted to node-positive patients (n = 304), comparable results were observed, in that VEGF121–206 and VEGF165–206 were statistically significant prognostic markers for OS as categorical variables [VEGF121–206, OS, P = 0.0038, RR, 2.09 (1.27–3.44); VEGF165–206, OS, P = 0.0105, RR, 2.22 (1.22–4.08)], and as continuous variables [VEGF121–206, OS, P = 0.0003, RR, 1.17 (1.08–1.28); VEGF165–206, OS, P = 0.0010, RR, 1.24 (1.08–1.42); Table 3⇓] . However, neither VEGF121–206 nor VEGF165–206 were significant markers for DFS in the node-positive cohort (Table 3)⇓ .
When univariate analysis was restricted to node-negative patients (n = 288), neither VEGF121–206 nor VEGF165–206 were of prognostic significance when tested as categorical or continuous variables for DFS or OS in the present study cohort with a follow-up duration of 52 months (Table 3)⇓ .
To demonstrate a biological concentration-effect relationship between VEGF expression levels and survival, node-positive patients with clinical follow up information (n = 304) were separated into three groups, including a low-risk group with VEGF121–206 levels below the 50th percentile (<676 fmol/mg), an intermediate risk group with VEGF121–206 levels between the 50th and 75th percentiles (676–1664 fmol/mg), and a high-risk group with VEGF121–206 levels above the 75th percentile (>1664 fmol/mg). Kaplan-Meier survival plots of each of these groups confirm that OS was associated with the absolute level of VEGF121–206 expression (log rank test P = 0.0097; Fig. 1A⇓ ). Comparable results were obtained with VEGF165–206 when node-positive patients were separated into three risk groups, with levels lower than the 50th percentile (<437 fmol/mg), between the 50th and 75th percentiles (437–835 fmol/mg), or above the 75th percentile (>835 fmol/mg; log rank test P = 0.0280; Fig. 1B⇓ ).
After univariate analysis, the variables of age, tumor size, number of positive axillary lymph nodes (0 versus 1–3 versus >3), hormone receptor status (ER- and/or PR-positive versus ER- and PR-negative), nuclear grade (1, 2 versus 3, 4), HER-2/neu status (positive versus negative), and VEGF121–206 or VEGF165–206 (as continuous variables following logarithmic transformation) were included in a multivariate Cox model. The analysis was performed among the entire cohort (n = 603) for DFS and OS, and among the node-positive patients (n = 304) for OS. When analyzing the entire cohort, number of positive axillary lymph nodes and steroid hormone receptor status were identified by the Cox multivariate regression model as statistically independent prognostic factors for DFS (Table 4)⇓ ; and age, number of positive axillary lymph nodes, and hormone receptor status were selected as statistically independent prognostic factors for OS (Table 4)⇓ . VEGF121–206 or VEGF165–206 did not retain independent prognostic significance for either DFS or OS when the entire cohort was analyzed, using these other parameters. When multivariate analysis for OS was restricted to the lymph node-positive subset of patients, age, nuclear grade, tumor size, number of positive lymph nodes, hormone receptor status, and VEGF121–206 or VEGF165–206 retained independent prognostic significance (VEGF121–206 P = 0.0103, RR 1.12 (1.04–1.22); VEGF165–206 P = 0.0150, RR 1.18 (1.03–1.34); Table 5⇓ ).
Combined Effect of HER-2/neu and VEGF.
Patients were next classified into four groups according to their HER-2/neu and VEGF status, including a low-risk group with normal HER-2/neu expression levels and no detectable expression of VEGF121–206, two intermediate groups with either HER-2/neu overexpression or detectable VEGF121–206 expression, and a high-risk group of patients with both HER-2/neu overexpression and detectable VEGF121–206 expression. When node-positive patients (n = 304) were analyzed, univariate analysis demonstrated statistically significant differences in survival between the four groups (log rank test P = 0.0092; Fig. 2A⇓ ), with the poorest outcome being among patients whose tumors demonstrated both HER-2/neu overexpression and VEGF121–206 expression. Conversely, the most favorable outcome occurred in patients whose tumors both demonstrated normal HER-2/neu expression and no detectable VEGF121–206 expression (Fig. 2A)⇓ . When all of the patients (n = 603) were separated into the same risk group, similar results were obtained; however, the differences in survival between these risk groups only reached borderline significance for the entire population (P = 0.0566; data not shown). Subsequently, we performed the same analysis, using VEGF165–206 levels in combination with HER-2/neu status, and the results were comparable with those obtained with VEGF121–206 (Fig. 2B)⇓ .
VEGF Isoform Patterns and Prognosis.
When analyzing the concordance of VEGF121–206 and VEGF165–206 expression, 321 (52.5%) samples tested positive in both VEGF121–206 and VEGF165–206 assays, and 120 (19.6%) samples tested negative for both VEGF isoforms. VEGF121–206 without VEGF165–206 protein was detected in 38 samples (6.2%), whereas 132 (21.6%) samples tested positive for VEGF165–206 and negative for VEGF121–206. Because the VEGF165 isoform also contains the lower amino acid sequences of VEGF121, a tumor that tests positive for VEGF165 should theoretically also test positive for VEGF121. The discrepancy in test results among the 132 samples of the latter group, however, could be based on methodological limitations of the detection methods. This may be due to the better sensitivity of the VEGF165–206 ELISA. Or, the epitope recognized by 2E3, the detecting antibody used in the VEGF121–206 ELISA, is unavailable in some patients for unknown reasons. Among the 321 samples in which both isoforms were detectable, VEGF121–206 positively correlated with VEGF165–206, as indicated in Fig. 3⇓ (r = 0.64, P < 0.001); however, there were substantial variations in the ratios between both isoforms. To investigate whether different VEGF isoform patterns could be associated with disease outcome, the VEGF121–206:VEGF165–206 ratios were calculated. Previous studies have shown that among the six VEGF isoforms, VEGF121 and VEGF165 are dominant in breast cancer cells, followed by isoforms 189, 183, and 145 at 10–100-fold lower concentrations (22) . The median VEGF121–206:VEGF165–206 ratio in the present study was 2.8 (range 0.3–48), confirming that the isoforms of 165 or more amino acids constitute approximately one-third of VEGF121–206 expression (35) . Using the median VEGF121–206:VEGF165–206 ratio as a cutoff point, patients with a ratio of >2.8 had DFS and OS comparable with patients with a ratio of ≤2.8 (data not shown), suggesting that different isoform patterns alone were not associated with prognosis in the current study cohort.
HER-2/neu overexpression is associated with increased tumor progression and metastasis (9 , 10 , 33) ; however, the exact mechanisms by which HER-2/neu regulates this more aggressive clinical phenotype are not fully understood. Recent studies indicate that HER-2/neu receptors play an important role in heregulin-induced angiogenesis (14, 15, 16) . Overexpression of the HER-2/neu receptor alone results in induction of the basal level of VEGF, and exposure to heregulin β1 additionally enhances VEGF secretion in breast cancer cells (14 , 15) . These experimental data indicate that an important consequence of HER-2/neu signaling is increased VEGF expression. VEGF in turn is a central regulator of angiogenesis, suggesting that the aggressive phenotype of HER-2/neu-overexpressing breast cancers may be in part attributable to increased angiogenesis.
In the present clinical study, a significant association between HER-2/neu overexpression and VEGF expression was demonstrated, with 77.2% of patients with HER-2/neu-overexpressing tumors demonstrating detectable VEGF121–206 expression, as opposed to 54.5% of their HER-2/neu-nonoverexpressing counterparts (P < 0.001). Importantly, this positive association was not sensitive to the cutoff chosen for HER-2/neu overexpression when measuring HER-2/neu protein by ELISA. Similarly, 87.7% of the patients with HER-2/neu-overexpressing tumors showed VEGF165–206 expression, as opposed to 71.0% of patients with HER-2/neu-nonoverexpressing tumors (P < 0.001). These clinical data confirm the preclinical experimental evidence indicating that HER-2/neu signaling can enhance the ability of a tumor to recruit a neovascular blood supply after the induction of VEGF expression (14, 15, 16) .
A recent study provides a molecular basis for the association between HER-2/neu overexpression and VEGF expression (36) . VEGF gene transcription can be mediated by hypoxia-inducible factor (HIF-1), which is a heterodimeric transcription factor (37) . HIF-1 activity is increased by both intratumoral hypoxia and genetic alterations, including loss of function mutations in tumor suppressor genes, as well as gain of functional alterations in oncogenes that activate the mitogen-activated protein kinase signal transduction pathways and the phosphatidylinositol 3′-kinase pathway (6 , 8 , 36 , 38 , 39) . HIF-1 is degraded rapidly under normal conditions by the ubiquitin pathway (40) ; however, under hypoxic conditions, HIF-1 is stabilized, and expression increases as a result of decreased ubiquination and degradation (36) . In contrast, heregulin stimulation does not affect the half-life of HIF-1, but instead stimulates HIF-1 protein synthesis at the transcriptional level (36) . Both mechanisms, hypoxia-driven inhibited degradation and HER-2/neu-stimulated increased synthesis of HIF-1, can therefore independently contribute to the up-regulation of VEGF (36 , 37) . The fact that HIF-1 can be activated by multiple genetic alterations, as well as hypoxia, is consistent with the observation in the current study that VEGF expression is associated with tumor characteristics that reflect these influences, such as hormone receptor levels, nuclear grade, histology, and tumor size. It is likely that tumor size contributes at least in part to elevated VEGF levels, because hypoxia is commonly observed in the microenvironment of breast cancers (41) . The fact that VEGF expression is subject to a number of other regulatory mechanisms can explain why a large proportion of patients with HER-2/neu-nonoverexpressing tumors do express VEGF (VEGF121–206, 54% and VEGF165–206, 71%), and why there is no correlation between the absolute levels of HER-2/neu and VEGF expression among patients with HER-2/neu-overexpressing tumors and detectable VEGF expression.
To date, preliminary clinical data have been reported on the association between HER-2/neu overexpression and VEGF expression in breast cancer, with two studies demonstrating a positive association between these two factors among 107 and 654 primary breast cancer patients (42 , 43) , and two other studies finding no such association among 123 (44) and 222 primary breast cancer patients (45) . The current study of 611 unselected primary breast cancer patients with 50-month median clinical follow-up, however, clearly supports the observation that HER-2/neu overexpression is associated with both VEGF121–206 and VEGF165–206 expression, and concurs with the results of Linderholm et al. (43) , who similarly studied a large number (n = 654) of unselected primary breast cancer patients.
Numerous studies have confirmed the clinical relevance of VEGF expression, regardless of HER-2/neu status, in breast tumors (46, 47, 48, 49, 50, 51, 52) . Our results on the prognostic relevance of VEGF121–206 are consistent with previous reports on unselected primary breast cancer patients (46 , 49) , in that VEGF121–206 is a significant prognostic marker for DFS and OS in univariate analysis. Similarly, our results confirm reports in which VEGF165–206 is found to be a significant prognostic marker for OS in univariate analysis (51) . We were, however, not able to confirm prognostic relevance of VEGF121–206 or VEGF165–206 in node-negative patients, as described previously (47 , 50) . In these earlier studies, however, node-negative patients had not received adjuvant treatment, whereas adjuvant therapy was commonly administered to high-risk node-negative patients in the current study cohort. The node-negative subset in our study may be underpowered to detect a significant difference in outcome, because among the 290 node-negative patients only 60 and 36 events occurred, respectively, during the 52 months of follow up for comparison of DFS and OS. In the current study VEGF121–206 and VEGF165–206 expression demonstrate prognostic independence for OS among the node-positive patients, which concurs with a previous study that similarly indicated that VEGF is an independent predictor of OS, using median values as cut-offs for elevated expression (51) . Moreover, when looking for a potential biological concentration response effect between VEGF expression levels and clinical outcome, we were able to demonstrate that absolute levels of VEGF121–206 and VEGF165–206 expression were correlated with survival in node-positive patients. This raises the possibility that VEGF levels might also be of importance as a predictive marker for therapeutic strategies that target VEGF (53) .
The circumstance that VEGF121–206 and VEGF165–206 were not detectable in 41.2% and 25.9% of the 611 invasive primary breast cancer samples could be explained by the possibility that the sensitivity of both assays is not sufficient to detect very low levels of VEGF expression. The absence of detectable VEGF expression in vascularized primary breast tumors, however, could also be explained by the fact that VEGF is not the only proangiogenic factor expressed in human breast cancers (2) . Platelet-derived endothelial growth factor, placenta growth factor, acidic and basic fibroblast growth factor, transforming growth factor α, transforming growth factor β, hepatocyte growth factor, and pleiotrophin have also been shown to be present during breast cancer progression (2) , and any single or combination of these molecules is capable of inducing angiogenesis.
In summary, the current study provides clinical evidence that HER-2/neu overexpression is associated with expression of two of the most abundantly expressed VEGF isoforms in breast cancer, suggesting that VEGF may in part mediate the aggressive phenotype of breast cancers that overexpress HER-2/neu. These data additionally support the use of combination therapies directed against both HER-2/neu and VEGF for treatment of breast cancers that contain the HER-2/neu alteration.
We thank Wendy Aft for invaluable assistance in preparing the manuscript.
Grant support: Department of Defense Breast Cancer Research Program (DAMD 17–01-1–0181).
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.
Requests for reprints: Gottfried E. Konecny, Department of Medicine, Division of Hematology-Oncology, David Geffen School of Medicine, University of California Los Angeles, 12–145 Factor Building, Le Conte Avenue, Los Angeles, CA 90095-1678.
- Received June 17, 2003.
- Revision received December 9, 2003.
- Accepted December 16, 2003.