Coexpression of the type 1 growth factor receptor family members HER-1, HER-2, and HER-3 has a synergistic negative prognostic effect on breast carcinoma survival

Abstract

BACKGROUND

The clinical significance of coexpression of type 1 growth factor receptor (T1GFR) family members remains largely unknown. The objective of the current study was to determine the frequency and the possible prognostic effect of coexpression of HER-1, HER-2, HER-3, and HER-4 by breast carcinoma.

METHODS

Tissue microarrays were constructed using clinically annotated formalin-fixed, paraffin-embedded tumor samples from 242 patients with invasive breast carcinomas with a median 15-year follow-up. The levels of TIGFR family members (HER-1–HER-4) were measured by immunohistochemistry. K-means clustering algorithm, as well as univariate (Kaplan–Meier, log-rank test) and multivariate (Cox regression) survival analyses were applied to the data set.

RESULTS

Using univariate analysis, expression of HER-1, HER-2, and HER-3, but not HER-4, was significantly associated with decreased patient disease-specific survival (P < 0.05). Kaplan–Meier survival analysis showed that coexpression of ≥ 2 of HER-1, HER-2, and HER-3 in any combination was associated with reduced patient disease-specific survival compared with single marker expression or no expression (35% vs. 65% vs. 78% 10-year survival rates, P = 0.001). Using multivariate analysis, expression of ≥ 2 of HER-1, HER-2, and HER-3 was independent of lymph node status and tumor size.

CONCLUSIONS

In a cohort of patients with breast carcinoma, the authors observed T1GFR family member coexpression (HER-1, HER-2, and HER-3) to have a negative synergistic effect on patient outcome, independent of tumor size or lymph node status. Thus, coexpression of T1GFR family members identified a subset of patients with a poor disease prognosis who may potentially benefit from therapy simultaneously targeting several T1GFR family members. Cancer 2005. © 2005 American Cancer Society.

Breast carcinoma is a leading cause of cancer-related mortality among women. The identification of clinicopathologic and molecular characteristics that allow an accurate prediction of a tumor's response to various treatments represents a critical cornerstone in the current management of this malignancy. Recognized disease prognosticators for individuals diagnosed with breast carcinoma include tumor size, histologic grade, and the presence of lymph node or distant metastases.1 These disease prognosticators alone, or in combination, enable the identification of individuals who are at increased risk of dying of their disease and also who may benefit from aggressive treatment (multimodal therapy). Unfortunately, breast carcinoma prognosticators do not provide the clinician with any guidance in selecting the specific therapy that would be of greatest benefit to an affected individual. Breast tumor expression of estrogen and progestin receptors has long played an important role in dictating whether hormonal therapy will be utilized as part of disease treatment.2 Recently, there has been much interest in the prognostic and treatment implications of the expression by breast tumors of the HER-2 receptor, a member of the type 1 growth factor receptor (T1GFR) family.3–6

The T1GFR family consists of 4 known transmembrane receptors: HER-1 (also known as epidermal growth factor receptor and c-erb B-1), HER-2 (c-erb B-2), HER-3 (c-erb B-3), and HER-4 (c-erb B-4). Signaling generated by members of the T1GFR family has evolved in a manner that is tightly controlled. The T1GFR receptors exist as either monomers or dimers (homodimers or heterodimers). HER-2 and HER-3 are only active when they form a heterodimer because HER-2 is a ligandless receptor and HER-3 lacks tyrosine kinase activity.3–6 The HER-2/HER-3 heterodimer has been described as the “pinnacle of ErbB receptor evolution” because it represents the development of a very powerful signaling module from a pair of singly inactive individual proteins.3 At least 30 ligands have been identified that bind T1GFR family members.7 Ligand binding to T1GFR family members leads to rapid receptor dimerization and it is the specific HER receptors expressed by breast epithelial cells that influences the dimers that form. Although there are multiple possible T1GFR homodimer and heterodimer combinations, it is HER-2 that preferentially heterodimerizes with other family members.8 Dimerization leads to autophosphorylation of tyrosine residues on the receptor's cytoplasmic domain and activation of signaling pathways involved in cell proliferation, survival, and transformation.3–6 The intracellular signaling generated by HER-2-containing heterodimers is the most potent of the entire T1GFR family in terms of cell growth and transformation.9 Despite HER-2 having no HER-activating ligand, its kinase not only potentiates signaling of HER-2-containing heterodimers, but also increases the binding affinity of ligands to HER-1 and HER-3.3–6 In a transgenic mouse breast carcinoma model, Siegel et al.10 demonstrated that coexpression of HER-2 and HER-3 transcripts leads to tumor progression and the development of distant metastases. Alimandi et al.11 showed evidence for cooperative signaling of HER-2 and HER-3 in neoplastic transformation of human breast carcinoma. In a cohort of 74 patients with invasive breast carcinoma, Hudelist et al.12 demonstrated that HER-2 and HER-3 were significantly (P < 0.001) coexpressed by the breast tumors of individuals with lymph node metastases. Thus, the HER-2/HER-3 heterodimers play an especially critical role in breast tumorigenesis and disease progression.

Members of the T1GFR family are overexpressed by human breast tumors. HER-2 is amplified or overexpressed in 10–25% of invasive breast carcinomas and gene amplification or overexpression predicts a poor patient prognosis.13–16 Women with tumors that overexpress HER-2 have a higher incidence of lymph node metastases, shortened time to disease recurrence, and shorter survival when compared with women whose tumors do not overexpress this receptor.13–16 HER-2 overexpression is also associated with a poorer response to tamoxifen and a better response to specific chemotherapeutic agents than would be expected.13–16 Thus, HER-2 expression has been useful in planning adjuvant breast carcinoma treatment. More recently, HER-2 has also served as a molecular target for drugs such as trastuzumab (Herceptin, Genentech, Inc., South San Francisco, CA), a monoclonal antibody that binds HER-2, shuts down signaling, and induces tumor regression.17

Although the HER-2 receptor is the best-studied T1GFR family member, a study of the pattern of expression of the entire family and of their relation to each other and to known breast carcinoma prognosticators, as well as their prognostic significance, would be of considerable clinical value. Thus, the objective of the current study was to determine the association of HER-1, HER-2, HER-3, and HER-4 expression with known clinicopathologic breast carcinoma prognosticators (tumor size, lymph node status, tumor grade) and patient outcome to determine whether they have any clinical value in the management of this patient population.

MATERIALS AND METHODS

Sequential archival samples from patients with invasive breast carcinoma, with available paraffin blocks, who had undergone treatment at Vancouver General Hospital between 1974 and 1995 were identified for tissue microarray (TMA) construction. The study was approved by the Clinical Research Ethics Board of the University of British Columbia. Hematoxylin and eosin-stained (H & E) sections of each tumor sample were examined and areas of invasive carcinoma marked on both the slide and the corresponding paraffin block for TMA construction. Adequate tissue specimens from all 242 patients with invasive breast carcinoma were available for immunohistochemical staining for HER-1, HER-2, HER-3, and HER-4.

TMA construction was performed with a tissue-arraying instrument (Beecher Instruments, Silver Springs, MD) using a method that was described previously.18 In brief, utilizing the arraying instrument, two 0.6-mm diameter cores of tumor specimen were removed from the marked region of the paraffin block, and transferred to defined array coordinates in a recipient TMA block. Three high-density TMA blocks were created for the current study. Utilizing a Leica microtome, serial 4-μm sections were cut and transferred to adhesive-coated slides. One section from each TMA block was stained with H&E and the remaining sections were stored for ≤ 3 weeks at room temperature before immunostaining was carried out. The antibodies and the antigen retrieval methodologies are summarized in Table 1.

Table 1. Characteristics of Antibodies Utilized for Immunohistochemistry

Antibody	Isotype	Clone	Company	Catalog no.	AgR	Concentration
HER-1	Mouse monoclonal	2-18C9	Dako	K1492	Proteinase K	Ready to use
HER-2	Rabbit polyclonal	2-18C9	Dako	A485	Steam 20 min, TRS	1 in 500
HER-3	Rabbit polyclonal	2-18C9	NeoMarkers	RB-066-PO	N	1 in 200
HER-4	Mouse monoclonal	HFR1	NeoMarkers	MS-637-PO	N	1 in 160

The H & E-stained sections and all immunostained slides were scanned with a BLISS automated digital imaging microscope (Bacus Laboratories, Lombard IL) and a digital archive of all pathologic data was created for scoring. The scoring of the TMA sections stained for estrogen receptor (ER), progesterone receptor (PR), HER-1, HER-2, HER-3, and HER-4 was performed in a semiquantitative manner. ER and PR scoring was accomplished using the method of Reiner et al.19 HER-1 and HER-4 were scored positive if any (weak or strong) cytoplasmic and/or membranous invasive carcinoma cell staining was observed. HER-2 and HER-3 staining was scored utilizing a 3-point scoring system: 0, invasive tumor cells present in the tissue core but no staining observed; 1, invasive tumor cells present with weak staining intensity and/or < 20% of tumor cells stained; 2, invasive tumor cells present with strong staining in > 20% of tumor cells. For the purpose of statistical analysis, score data were binarized, using the cutoffs of ≥ 10% for ER and PR-positive staining, a score of 1 or 2 represented positive staining for HER-1 and HER-4, and a score of 2 (equivalent to strong 3+ staining when utilizing the HercepTest®[Dako Corporation, Carpinteria, CA]) represented positive staining for HER-2 and HER-3 (Fig. 1). All samples were evaluated and scored from a scanned image on a computer monitor simultaneously by three pathologists who were blind to patient clinical data. Any interpathologist discrepancies in the scoring of a specific tissue core were immediately resolved with the majority score, or the score given by two of the three pathologists, being assigned. If there was a discrepancy in the scores assigned to the 2 cores from the same tumor, the tumor was assigned the higher score.

Figure 1. Tissue cores exhibiting strong expression of HER family members. Strong membranous staining for (A) HER-1 and (B) HER-2. Cytoplasmic staining for (C) HER-3 and (D) HER-4). All cores were scanned using a × 20 objective lens.

All data was logged into a standardized score sheet matching each TMA section (Microsoft Excel; Microsoft, Redmond, WA). The spreadsheet was then processed utilizing TMA-Deconvoluter 1.06 software that had been adapted for TMA analysis.20 The data were then analyzed by the SPSS statistical software package (Version 11; SPSS, Chicago, IL). Correlation analysis was performed utilizing the bivariate two-tailed Spearman nonparametric correlation test. Differences were considered significant at P < 0.05. A K-means clustering analysis algorithm was applied to identify tissue specimens with different combinations of positive immunostains for HER-1, HER-2, HER-3, and HER-4. The maximum number of 15 clusters was predefined before the clustering procedure. After the clustering was done, for the purpose of statistical analysis, the initial clusters were merged according to the number of markers (HER-1, HER-2, HER-3, or HER-4) coexpressed simultaneously (none, 1, 2, ≥ 3). The Kaplan–Meier method was utilized for survival analysis and the log-rank test was utilized to determine the statistical significance of differences between survival curves. The Cox proportional hazard regression model was utilized to carry out multivariate survival analysis.

Clinical data for all patients were retrospectively collected from patient hospital charts. Median patient follow-up was 15 years. All patients had newly diagnosed Stage I–III invasive breast carcinoma. Clinicopathologic data included patient age, gender, lymph node status (negative vs. positive), tumor size, tumor grade, tumor histology, patient follow-up, and survival.

RESULTS

Correlational Analysis

No correlations were found between T1GFR family members themselves, except for a strong positive correlation between HER-3 and HER-4. HER-1 and HER-2, but not HER-3 or HER-4, showed significant correlations with ER and PR negative status of tumor specimens. HER-1 and HER-2 also showed significant correlations with tumor grade. None of the T1GFR family members showed significant correlations with tumor size or axillary lymph node status (Table 2).

Table 2. Correlation Table for a Cohort of 242 Patients with Invasive Breast Carcinoma

Marker Spearman rho	HER-1	HER-2	HER-3	HER-4
PR: progestin receptor; ER: estrogen receptor. a Correlation is significant at the 0.01 level (two-tailed).
HER-2
Correlation coefficient	0.044
P value (2-tailed)	0.448
HER-3
Correlation coefficient	0.05	−0.051
P value (2-tailed)	0.423	0.409
HER-4
Correlation coefficient	0.037	−0.04	0.431a
P value (2-tailed)	0.545	0.507	0
PR
Correlation coefficient	−0.393a	−0.211a	0.093	0.078
P value (2-tailed)	< 0.001	< 0.001	0.13	0.197
ER
Correlation coefficient	−0.620a	−0.292a	0.028	0.035
P value (2-tailed)	< 0.001	< 0.001	0.66	581
Grade
Correlation coefficient	0.363a	0.231a	−0.031	−0.007
P value (2-tailed)	< 0.001	< 0.001	0.607	0.903
Lymph node status
Correlation coefficient	−0.073	−0.019	0.063	0.069
P value (2-tailed)	0.198	0.741	0.325	0.27
Tumor size
Correlation coefficient	0.114	−0.022	0.004	0.03
P value (2-tailed)	0.054	0.705	0.951	0.636

Univariate Analysis of Disease-Specific Survival

A higher level of expression of HER-1, HER-2, and HER-3, but not HER-4, was associated with a decreased disease-specific survival (Fig. 2). For this reason HER-4 was excluded from further coexpression and survival analyses.

Figure 2. Kaplan–Meier survival curves for a cohort of 242 patients with invasive breast carcinoma demonstrating tumor expression of (A) HER-1, (B) HER-2, (C) HER-3, and (D) HER-4.

Coexpression of T1GFR Family Members and Effect on Disease-Specific Survival

A single marker was expressed (HER-1, HER-2, or HER-3) in 68 (28.1%) patients. In 11 patients (4.6%), there was coexpression of ≥ 2 T1GFR family members. In 163 (67.4%) patients, there was negative expression of all 3 T1GFR family members. There was a distinct distribution of events (breast carcinoma-related deaths) in each group when compared with each other, and when compared with the group of patients with negative expression of all three T1GFR family members (Table 3). Disease-specific survival analysis of HER family member expression and coexpression showed decreased disease-specific survival depending on the expression and coexpression of ≥ 2 T1GFR family members by the same tumor. The highest 20-year survival rate was found in a group of patients whose tumor specimens were negative for all 4 T1GFR family members (75.5%), intermediate in a group of patients whose tumor specimens expressed 1 T1GFR family member (60.3%), and worst in a group of patients whose tumor specimens coexpressed ≥ 2 T1GFR family members (36.4%) (Fig. 3A). A trend toward worse survival for coexpressors of two or more T1GFR family members was also observed for overall patient survival (Fig. 3B).

Table 3. Expression and Coexpression of HER-1, HER-2, and HER-3 in a Cohort of 242 Patients with Invasive Breast Carcinoma with Distribution of Events by Group and Marker Combinationa

Combination	Markers	No. of patients (%)	Death from breast carcinoma	Death from breast carcinoma (summary by group)
a Type 1 growth factor receptor family coexpressors are shown in bold type.
Single marker expression	HER-1	22 (9.1%)	8 (36.4%)	27/68 (39.7%)
68 (28.1%)	HER-2	24 (9.9%)	10 (41.7%)
	HER-3	22 (9.1%)	9 (40.1%)
Two or more markers	HER-1 + HER-2	4 (1.7%)	2 (50%)	7/11 (63.6%)
Coexpression	HER-1 + HER-3	4 (1.7%)	3 (75%)
11 (4.6%)	HER-2 + HER-3	2 (0.8%)	1 (50%)
	HER-1 + HER-2 + HER-3	1 (0.4%)	1 (100%)
All negative		163 (67.4%)	40 (24.5%)	40/163 (24.5%)
Total		242 (100%)	74 (30.5%)	74/248 (29.8%)

Figure 3. Kaplan–Meier survival curves showing (A) disease-specific survival and (B) overall survival for nonexpression, single-receptor expression, and coexpression of HER-1and/or HER-2 and/or HER-3.

Multivariate Analysis

Only T1GFR family members and other clinicopathologic variables that showed statistical significance (P < 0.05) by univariate analysis were entered into the model for further analysis. In the first model, only HER-1, HER-2, and HER-3 were considered. HER-2 and HER-3 showed an independent effect on disease-specific survival, whereas HER-1 approached significance (P = 0.079; 99% confidence interval). There was no statistically significant interaction between markers in the model when the interaction test was performed.

In the second model, HER-1, HER-2, and HER-3 were considered along with ER, tumor grade, tumor size, and lymph node status. Of the 7 variables evaluated in the model, only HER-3 (P = 0.011) and lymph node status (P = 0.003) were found to be of independent prognostic significance.

The categorical variables entered into the third model included negative expressors, single T1GFR family member expressors, and coexpressors of two or more T1GFR family members, along with patient lymph node status (lymph node positive vs. lymph node negative) and tumor size (> 20 mm or ≤ 20 mm in greatest dimension). In this model, the prognostic significance of the overall expression of T1GFR family members was independent of patient lymph node status and tumor size (P = 0.011). Coexpression of ≥ 2 T1GFR family members was associated with a 3.7 times higher risk of death when compared with negative HER receptor status, and was also independent of lymph node status and tumor size (P = 0.003). Singular expression of either HER-1, HER-2, or HER-3 was associated with a 1.5 times higher risk of death when compared with nonexpressors, but this effect was not independent of lymph node status and tumor size (P = 0.174). In this model, the presence of lymph node metastasis was an independent prognostic marker (P = 0.002; relative risk [RR] = 2.5) and tumor size showed a trend toward independent prognostic significance (P =0.062; RR = 1.7).

DISCUSSION

Overexpression of the T1GFR family members HER-1 and HER-2, by human breast tumors, predicts a poor patient prognosis.13–16 Slamon et al.21 were the first to describe HER-2 gene amplification, utilizing a Southern blot analysis technique, as predicting overall survival and time to disease recurrence in a cohort of 189 patients with breast tumors. They found that in patients with lymph node disease, HER-2 amplification was of greater prognostic value than hormone receptor status. Other groups have also found that overexpression of the HER-2 receptor by patients with breast carcinoma also was predictive of a poor disease prognosis.13, 22, 23 Multiple investigators have also found HER-1 overexpression to represent a negative prognosticator for patients with breast carcinoma.24, 25 However, to our knowledge, few groups to date have evaluated the expression of the entire T1GFR family in human breast carcinoma. Utilizing immunohistochemistry, Witton et al.26 evaluated T1GFR family expression in tissue specimens from a cohort of 220 patients with breast carcinoma treated over a 9-year period in Glasgow, Scotland. In their study, overexpression of a single T1GFR family member was observed in 48.6% of tissue specimens, multiple family members in 16.4% of tissue specimens, and HER-2 and HER-3 overexpression was highly correlated (P < 0.0000001). In a cohort of 35 patients with lymph node positive breast carcinoma, Esteva at al.27 reported an association between HER-1, HER-2, and HER-3 coexpression. In the current study, we observed lower rates of T1GFR family member overexpression, with a single member being overexpressed in 28.1% of tissue specimens, and multiple members overexpressed in 4.6% of tissue specimens. We did not observe any correlation between HER-2 and HER-3 overexpression. However, in our study, we did observe a strong correlation between HER-3 and HER-4 receptor expression (P < 0.01), much higher than the result reported by Witton et al.,26 who observed overexpression of both of these receptors in only 1.4% of their tissue specimens. In a cohort of 66 patients with lymph node positive breast carcinoma evaluated by immunohistochemistry for T1GFR family expression, Lodge et al.28 also reported a strong association between HER-3 and HER-4 overexpression (P = 0.003). Similar to Witton et al.,26 we were able to demonstrate decreased survival of individuals whose tumor specimens overexpressed either HER-1, HER-2, or HER-3. We were unable to demonstrate any relation between HER-4 expression and survival, whereas Witton et al.26 observed improved survival with HER-4 expression. Lodge et al.28 found HER-4 expression to predict poor patient prognosis. The prognostic significance of HER-4 expression by breast carcinoma is currently unclear and has been conflicting in numerous other reports.29, 30 Similar to Witton et al.,26 we also demonstrated HER-1 and HER-2 overexpression to be significantly associated with tumor grade and negative ER expression. Suo et al.31 evaluated a Swedish cohort of 100 patients with invasive breast carcinoma utilizing immunohistochemistry and reverse transcription-polymerase chain reaction. They observed that HER-2 overexpression was associated with a significantly shorter disease-free survivial (P = 0.033) and cancer-specific survival (P = 0.042). Conversely, HER-4 overexpression was associated with a significantly longer disease-free survival (P = 0.049) and cancer-specific survival (P = 0.044). In this Swedish cohort, HER-1 and HER-3 overexpression did not correlate with patient outcome. This group also found cooverexpression of HER-1 and HER-2 to be associated with a worse patient prognosis. HER-4 overexpression was observed to have an antagonistic effect on the worse outcomes in patients with HER-2 overexpression. Suo et al.32 also described T1GFR family expression in a separate cohort of 107 patients with invasive breast carcinoma treated in China. In this Chinese cohort, HER-2 overexpression was weakly associated with poor patient prognosis and HER-4 overexpression was associated with a good prognosis, but only in the absence of lymph node disease. In a TMA study of 324 patients with lymph-node negative invasive breast carcinoma, Ocal et al.33 were unable to demonstrate that HER-1 or HER-2 overexpression had any prognostic value whatsoever. Unlike these earlier studies evaluating T1GFR family expression by human breast carcinoma, we found that HER-1, HER-2, and HER-3 each independently worsened patient prognosis—the greater the number of these family members that are overexpressed by the tumor, the poorer the patient prognosis. Utilization of a TMA to evaluate T1GFR family expression by breast tumors appears more clinically relevant than previous whole-section studies, in which overestimation of focal immunopositivity can lead to devaluation of the prognostic value of important molecular markers.

To our knowledge, the current study is the first to describe the negative synergistic effect on the outcome of patients with breast carcinoma of coexpression of the T1GFR family members HER-1, HER-2, and HER-3. These results underscore the complexity of the signaling generated by the T1GFR family. Signaling generated by the T1GFR family must not be viewed as a collection of linear pathways, but rather as a complex and highly integrated system with crosstalk, networking, and redundancy occurring between multiple important effectors.34 These results also have therapeutic implications for the management of patients with breast carcinoma. The monoclonal antibody trastuzumab has been utilized successfully as monotherapy, and in combination with chemotherapy, in women diagnosed with HER-2—overexpressing metastatic breast carcinoma.35, 36 Preclinical studies have demonstrated synergistic effects of HER-1 and HER-2 inhibitors in HER-2–overexpressing tumors.37, 38 Numerous clinical trials are currently underway in cohorts of patients with breast carcinoma to investigate strategies that target multiple T1GFR family members by utilizing combinations of T1GFR inhibitors or new agents capable of blocking multiple family members.39, 40 Further evaluation of the relation of T1GFR family expression with specific treatments utilized may potentially improve the predictive value and the clinical applicability of these markers. Thus, the prognostic and potential treatment implications of tumor expression of HER-1, HER-2, and HER-3 are significant, warrant further study, and may ultimately provide insights into disease biology that will benefit women diagnosed with breast carcinoma.