Introduction
One of the biggest challenges in today’s breast cancer care is to adjust adjuvant treatment, according to both tumor and patient characteristics, for optimal treatment aimed at improved prognosis. Although both tamoxifen and aromatase inhibitors have been shown to improve survival in estrogen receptor-positive breast cancer, the disease will recur in many patients despite adjuvant treatment (7–9% breast cancer recurrences five years after randomization in BIG-98 [
1]). Moreover, in the metastatic setting most tumors eventually develop resistance to the given treatment. Hence, further studies of potential prognostic and predictive biomarkers are essential in order to optimize and individualize breast cancer treatment.
An interesting biomarker in relation to endocrine treatment is AIB1 (amplified in breast cancer 1). AIB1 belongs to the p160 steroid receptor coactivator family and interacts with the estrogen receptor in a ligand-dependent manner to enhance transcription [
2,
3]. However, it has also been shown to interact with other transcription factors and signaling pathways inducing hormone-independent proliferation [
4‐
6]. In human breast cancer, AIB1 correlates with factors indicating a more aggressive phenotype (
HER2 amplification, DNA non-diploidy, a high tumor grade, a high S-phase fraction, and high Ki67) [
5,
7‐
10]. Several studies have also indicated AIB1 to be associated with endocrine treatment effect [
5,
7,
8,
11‐
13], although results have not been unanimous. We have previously shown AIB1 to be both a prognostic marker and a predictive marker for adjuvant tamoxifen in a randomized trial of premenopausal women receiving adjuvant tamoxifen for 2 years versus control, and in independent cohorts [
9,
10,
14]. These data were extended also to postmenopausal patients in an independent randomized trial of adjuvant tamoxifen versus control [
15]. Women with estrogen receptor-positive breast cancer expressing high levels of AIB1 have a worse prognosis, but respond well to tamoxifen. The prognosis of women with low tumor expression of AIB1, on the other hand, is not further improved by tamoxifen, although early on they have a better prognosis. However, previous studies of AIB1 in relation to aromatase inhibitors are very sparse, and its predictive value for treatment with aromatase inhibitors has not been evaluated in any larger clinical trial. If patients with low tumor expression of AIB1 would still benefit from aromatase inhibitors, AIB1 might be a predictive marker for the choice between tamoxifen and aromatase inhibitors, which is something we lack in the clinic today.
We use the Danish subcohort of the large randomized Breast International Group (BIG) 1-98 trial of adjuvant tamoxifen versus letrozole (as monotherapy or sequentially) with the aim to investigate AIB1 as a prognostic and predictive biomarker in relation to adjuvant endocrine treatment in estrogen receptor-positive postmenopausal breast cancer.
Patients and methods
BIG 1-98
The design of the BIG 1-98 trial, as well as the Danish cohort, has been described in detail before [
16‐
18]. Briefly, this is a randomized, phase 3, double-blinded trial of postmenopausal, estrogen receptor-positive, early breast cancer. Patients were randomized to either monotherapy with tamoxifen or letrozole for 5 years, or to sequential therapy with 2 years of tamoxifen or letrozole followed by an additional 3 years with the other drug (letrozole/tamoxifen). The trial enrolled 1396 Danish patients from 1998 to 2003 included in the intention-to-treat population. Primary tumor samples were available for 1323 of patients and tissue microarrays were constructed for 1281 of these [
18‐
20]. In 1997, the Danish Medicines Agency and the Danish National Committee on Biomedical Research Ethics approved the double-blinded BIG 1-98 trial, and the Ethical Committee of the Capital Region approved the current biomarker study before its activation (KF 02-178/97, KF 12-142/04, RH-2015-166; I-suite 04070). The BIG 1-98 trial is registered on the clinical trial site of the USA National Cancer Institute’s website (http:
www.clinicaltrials.gov/ct/show/NCT00004205). The remark criteria were considered for presentation of data below [
21].
Central assessment of the estrogen receptor, progesterone receptor, and HER2
The International Breast Cancer Study Group’s Central Pathology Laboratory carried out a central review on whole tissue sections for estrogen and progesterone receptors by immunohistochemistry, and for
HER2 by immunohistochemistry and fluorescent in situ hybridization as previously described [
1,
22]. Tumors expressing estrogen or progesterone receptors in ≥1% of tumor cells were considered hormone receptor positive, and those with a
HER2-to-Centromere-17 ratio ≥2 considered
HER2-amplified. The pathology central review was carried out without knowledge of other characteristics, treatment assignment, or outcomes.
Immunohistochemistry for AIB1
Tissue microarrays were constructed from formalin-fixed and paraffin-embedded tumor blocks by a tissue microarray builder using 2-mm tissue cores [
18]. Each tumor was represented by two cores. Immunohistochemistry for AIB1 was performed in an Autostainer-
Plus, Dako. As a primary antibody for AIB1 detection, a mouse-monoclonal IgG antibody was used at 1:100 dilution (Cat no #611105 BD Bioscience, CA, USA), as previously described [
7,
10]. This antibody has been used in several previous clinical trials [
3,
7,
8], and its specificity has been confirmed by both Western blot and Northern blot, and in situ hybridization [
3,
8]. Immunohistochemical staining (nuclear) was examined by two independent viewers blinded for clinical/tumor characteristics (Sara Alkner and Kristina Lövgren). Each sample was semi-quantitatively scored from 0 to 3 for percentage of stained nuclei and staining intensity. Proportion score 0 represented no stained nuclei, 1:1–10%, 2:11–50%, and 3:51–100%. Staining intensity 0 represented negative staining, 1 weak, 2 moderate, and 3 intense staining. Proportion and intensity scores were added to a total score ranging from 0 to 6. As in several previous publications from our group, a total score of ≥5 was used to define high AIB1 [
7,
9,
10,
14]. Cases classified differently (high vs. low AIB1) between viewers (8%) were reexamined to reach consensus. In case of discrepant staining between the two cores from the same patient, the highest score was used.
Statistical analysis
All clinical data were collected and monitored by the International Breast Cancer Study Group and subsequently transferred to the Danish Breast Cancer Cooperative Group, where the statistical analyses were conducted. Follow-up time was quantified in terms of a Kaplan–Meier estimate of potential follow-up. The primary end-point was disease-free survival, defined as the time from randomization to the first of the following events: recurrence at local, regional, or distant sites; a new invasive cancer in the contralateral breast; any second (non-breast) malignancy; or death without a prior cancer event. Secondary end-point was overall survival, defined as the time from randomization to death, irrespective of cause of death. Time-to-event outcomes were analyzed according to intention to treat. Follow-up was censored at last disease assessment, and in cross-over arms for predictive analysis at 2 years: time of scheduled cross-over.
Baseline characteristics were compared using two-sided Fisher’s exact tests or Wilcoxon rank sum test (age and tumor size). AIB1 expression (low/high) was compared via a stratified log-rank test of disease-free and overall survival, with randomization option (two-arm or four-arm) and treatment arm as a stratification factor, and Kaplan–Meier plots were generated. Cox proportional hazards models were used to estimate hazard ratios and 95% confidence intervals stratified by randomization option; multivariable models were adjusted for age at surgery, tumor size, tumor type and grade, estrogen receptor status, HER2 status, and nodal status. Assumptions of proportional hazard were tested using Schoenfeld residuals and by including a time-dependent component. The interactions of treatment by subpopulation of AIB1 were tested by Cox proportional hazards models including treatment groups, an indicator of the subpopulation, and the interaction term, and likewise for interaction of HER2 and the estrogen receptor by subpopulation of AIB1. Level of statistical significance was set to 5%. Statistical analyses were performed with the SAS v9.4 program package.
Discussion
We have investigated the estrogen receptor coactivator AIB1 as a prognostic and predictive factor in relation to endocrine treatment in the Danish cohort of BIG 1-98 (tamoxifen vs. letrozole). We found 46% of tumors to express high levels of AIB1, which is similar to previous publications [
9,
10]. In line with earlier studies, AIB1 correlated to a more aggressive tumor phenotype (
HER2 amplification and a high malignancy grade) [
5,
7,
8,
23,
24]. In relation to estrogen and progesterone receptor status, results from previous studies have varied. Some show a high AIB1 to be associated with hormone receptor-positive disease [
8,
25], while others report an association with estrogen and progesterone receptor negativity [
5,
23,
26], or no correlation to receptor status at all [
2,
7,
24]. Differences may possibly be explained by different study designs, cut-offs, and study populations. In this cohort, a high AIB1 was associated with a higher estrogen receptor expression.
AIB1 has been reported to be a negative prognostic factor, both in estrogen receptor-positive and estrogen receptor-negative breast cancers [
9‐
11,
14,
24,
26‐
29]. As a result of studies that show an inferior prognosis for AIB1-high tumors in endocrine-treated estrogen receptor-positive breast cancer cohorts [
5,
7,
8], AIB1 has previously been suggested to be of importance for endocrine treatment resistance. However, our earlier investigations of a randomized premenopausal estrogen receptor-positive breast cancer trial (2 years adjuvant tamoxifen vs. control) clearly indicate that although AIB1-high tumors had an inferior prognosis early on, these patients responded very well to tamoxifen [
10]. AIB1-low tumors had a better prognosis early on, but this was not further improved by tamoxifen. Importantly, these results have been confirmed in a randomized tamoxifen trial including postmenopausal estrogen receptor-positive disease [
15]. Hence, we hypothesize that the association between a high AIB1 expression and poor prognosis is related to its prognostic significance, which cannot be entirely eradicated by adjuvant endocrine treatment. The strong prognostic effect of AIB1 makes it a very interesting target for new anti-cancer therapies. Although steroid receptor coactivators are large unstructured proteins making production of drugs against them challenging, there are ongoing efforts to pharmacologically target them [
30,
31].
Today there are no markers that predict which population of postmenopausal estrogen receptor-positive breast cancer patients that are likely to have superior benefit from tamoxifen versus aromatase inhibitors. Due to AIB1’s predictive effect in relation to tamoxifen, we postulated that AIB1 status might be a useful marker. Previous trials regarding AIB1’s relation to aromatase inhibitors are very sparse, include small cohorts, and show conflicting results [
12,
13]. The data presented here are, to our knowledge, the first time AIB1 is investigated in relation to aromatase inhibitors in a large randomized clinical trial. However, we found no evidence of differences in treatment effect between tamoxifen and letrozole in relation to AIB1 status. Hence, our study indicates that tumor expression of AIB1 cannot be applied as a predictive marker for selection of tamoxifen versus letrozole as adjuvant therapy in postmenopausal endocrine-responsive breast cancer.
A relationship between AIB1 and HER2 has previously been suggested, with a worse prognosis with co-expression of AIB1 and HER2 [
5,
8,
32]. We found a correlation between a high AIB1 and
HER2 amplification. However, in line with our previous studies, no significant interaction between AIB1 and HER2 in relation to prognosis was detected [
9,
14]. As in all earlier studies though, AIB1-high,
HER2-amplified tumors represented only a small subgroup of the cohort.
Although we had the advantage of using a large international controlled randomized trial, this study still has some potential limitations. Most importantly, the BIG 1-98 trial did not include a control group of patients not receiving endocrine therapy. Access to such a group would probably have clarified the prognostic and predictive value of AIB1 even more, especially in relation to letrozole. In addition, after an interim analysis showed a superior effect for letrozole, patients randomized to tamoxifen were allowed a treatment switch, reducing the possibility to detect differences in treatment effect [
17]. Furthermore, although a large patient cohort is included, as in all studies, numbers are strongly reduced in subgroup analyses, such as investigations of AIB1-high/
HER2-amplified tumors. Finally, although we used a cut-off to define a high AIB1, which has been used in several previous studies, AIB1 is still an explorative biomarker and an optimal cut-off is yet to be definitely determined.
In conclusion, in a subset of the BIG 1-98 study population, we confirm tumor expression of AIB1 to be a strong negative prognostic factor. As the association with a high AIB1 and poor prognosis has now been repeatedly shown in different patient cohorts, AIB1 is an interesting target for anti-cancer therapies. However, no difference in treatment effect between tamoxifen and letrozole in relation to AIB1 was found. Hence, AIB1 cannot be of assistance for the choice of type of endocrine treatment in postmenopausal endocrine-responsive disease.
Acknowledgements
We thank the International Breast Cancer Study Group for providing data on the Danish patients enrolled in the BIG 1-98 trial. We also thank the patients, nurses, data managers, and physicians who contributed to the BIG 1-98 clinical trial; and the Danish Cancer Cooperative Group for collaboration on this project. Finally, we thank Kristina Lövgren for immunohistochemical staining and scoring of AIB1, and Simon Hayes for proofreading.