Androgen Receptor Is Expressed in the Majority of Breast Cancer Brain Metastases and Is Subtype-Dependent

Simple Summary Androgen receptor (AR) is a receptor found on many breast cancer cells, and drugs can be used to target AR in the treatment of advanced breast cancers. We aim to better understand how AR is expressed in breast cancer that has spread to the brain, i.e., brain metastases (BrM). We used a technique called immunohistochemistry to study AR expression in BrM from 57 breast cancer patients. We found that AR was expressed in the majority of BrM, and was expressed at different frequencies across different types of breast cancer. We did not find an association between AR expression and measures of survival. In most patients, AR expression in breast tumours and BrM was comparable. This study shows that AR is expressed in the majority of breast cancer BrM, and may be a useful target for treating breast cancer patients who have BrM. Abstract We aimed to evaluate the expression of the “targetable” androgen receptor (AR) in breast cancer brain metastases (BrM). An established, retrospective 57-patient cohort with metastatic breast cancer who underwent surgery for BrM at the Sunnybrook Odette Cancer Centre between 1999–2013 was studied. AR expression in BrM samples was assessed in triplicate using immunohistochemistry (IHC). AR positive status was defined as nuclear AR expression ≥ 10% by IHC using the SP107 antibody. The median age of patients was 52 years (range 32–85 years). 28 (49%) of BrM were HER2+, 17 (30%) were hormone receptor positive (HR+)/HER2−, and 12 (21%) were triple negative breast cancers (TNBCs). 56% (n = 32/57) of BrM were AR positive, and median AR expression was 20% (CI 1.6–38.3%). AR expression was different across breast cancer subtypes; AR was most frequently expressed in HER2+ (n = 21/28), followed by HR+/HER2− (n = 9/17), and lowest in TNBC (n = 2/12) BrM (p = 0.003). Patients with AR positive versus AR negative BrM had similar overall survival (12.5 vs. 7.9 months, p = 0.6), brain-specific progression-free survival (8.0 vs. 5.1 months, p = 0.95), and time from breast cancer diagnosis to BrM diagnosis (51 vs. 29 months, p = 0.16). AR is expressed in the majority of breast cancer BrM and represents a potential therapeutic target.


Introduction
Androgen receptor (AR) is a nuclear steroid hormone receptor that functions as a ligand-activated transcription factor and regulates the expression of downstream genes involved in cellular proliferation and apoptosis [1,2]. AR is ubiquitously expressed in 72-79% of primary breast cancers [3][4][5], although its expression differs by breast cancer subtype, and is more frequently expressed in hormone receptor positive breast cancer compared to hormone receptor negative breast cancer [5,6]. Triple-negative breast cancer (TNBC) is traditionally thought to be devoid of molecular targets, yet AR is expressed in 27-35% of cases [3,5,7]. As a result, several phase II trials have evaluated the safety and efficacy of androgen-targeted therapies, including abiraterone, enzalutamide and bicalutamide, among patients with metastatic TNBC with clinical benefit rates ranging from 19-35% [8][9][10][11][12]. The safety and efficacy of androgen-targeted therapies with standard of care systemic therapies, such as trastuzumab for patients with advanced HER2+ breast cancer [13], and exemestane for those with hormone receptor (HR) positive disease, have also been evaluated [14].
Given that breast cancer is the second most common cause of brain metastases (BrM), which are associated with high rates of morbidity and mortality, there is interest in evaluating novel treatment approaches with central nervous system (CNS) activity [15]. The anti-androgen agents enzalutamide and apalutamide are known to cross the blood-brainbarrier and hold promise as targeted treatment options for patients with AR-positive metastatic disease [16]. However, little is currently known about the expression of AR in breast cancer BrM [17].
We aimed to evaluate the expression of AR using a clinically validated and commercially available antibody in an established cohort of patients with metastatic breast cancer who underwent surgery for BrM at our institution.

Study Population
This study used a retrospective cohort of breast cancer patients who underwent surgery for BrM at Sunnybrook Odette Cancer Centre (SOCC) in Toronto, Ontario, Canada, consecutively between July 1999 and June 2013. The Anatomic Pathology Laboratory Information System was searched for metastatic carcinoma of breast origin. Patients with a personal history of previous or concurrent breast carcinoma and BrM and no evidence of another primary tumor were included if their breast primary pathology was available in the archives of the Pathology Department. Clinical information as well as information on BrM and matched primary breast tumours were obtained from the patients' electronic medical records. Mean followup time was 11.4 months. Research Ethics Board approval was obtained from the Sunnybrook Research Institute.

Immunohistochemistry
Immunohistochemistry (IHC) was assessed on BrM samples using an established tissue microarray (TMA) consisting of 1 um tissue cores in triplicates. AR expression was assessed using the SP107 antibody (Sigma-Aldrich (St. Louis, MO, USA)) and was scored as a continuous variable as percentage of positive nuclei. Positive status was defined as nuclear AR expression ≥10% of tumor cells. GATA3 expression was assessed using the sc-268 antibody (Santa Cruz Biotech (Dallas, TX, USA)) and Ki-67 expression was assessed using the clone 30-9 antibody (Ventana (Tucson, AZ, USA)). Both GATA3 and Ki-67 expression were scored as a continuous variable as percentage of positive nuclei, and we defined low expression as 1-24%, intermediate expression as 25-49%, and high expression as ≥50%. Estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor (HER2) status of the BrM was evaluated based on 2020 and 2018 ASCO/CAP guidelines respectively [18]. Five patients had two resected BrM; in these cases we took the average AR expression of the two BrM. AR expression was also evaluated in the matched primary breast tumours for 10 patients using similar methods.

Statistical Analysis
Brain-specific progression-free survival (bsPFS), OS, and time from diagnosis of breast cancer to diagnosis of BrM were estimated using the Kaplan-Meier method and compared across groups using the log-rank test. Independence between categorical variables was assessed using the χ2 test. Mean AR expression across breast cancer subtypes was compared using ANOVA, and post-hoc analysis using Tukey HSD. Difference between AR expression in primary breast samples and matched BrM was tested using the Wilcoxon signed rank test. bsPFS was defined as the duration of time in months from the time of BrM surgery to the time of BrM progression or death. Overall survival (OS) was defied as the duration of time in months from the time of BrM surgery to the time of death due to any cause. Time from diagnosis of breast cancer to diagnosis of BrM was defined as the duration of time in months from first diagnosis of primary breast cancer to the time of BrM surgery. For all analyses, statistical significance was defined as a p-value of <0.05. Data was analyzed using R version 4.2.1 (23 June 2022) and visualized using the ggplot2 package.

Systematized Review of Active Clinical Trials Investigating AR Targeted Therapies in Advanced and Metastatic Breast Cancer
A search was conducted on 25 January 2023 on the clinicaltrials.gov website using the search terms "Breast cancer" and "Androgen" and filtered for studies that were "Recruiting", "Not yet recruiting" or "Active not recruiting". This search yielded 37 trials, which were manually reviewed. Only trials including women ≥18 years with advanced or metastatic breast cancer, with the primary intervention being an androgen-targeted therapy either alone or in combination with other systemic therapies were selected for review. 16 trials met the inclusion criteria and are included in Table 1.
In subsequent analyses, cases with AR expression ≥10% were labelled as "AR positive (AR+)", while those with AR expression <10% were labelled as "AR negative (AR−)". This was based on a previous study that found that AR expression threshold of 10% is associated with clinical response to enzalutamide in TNBC [8]. 56% (n = 32/57) of BrM were AR positive. Distribution of BrM subtype, Ki-67 expression, GATA3 expression, age at BrM diagnosis, BrM location, BrM size, and BrM grade, as stratified by AR status, is shown in Table 3. No significant associations were found between AR status and age at BrM diagnosis, BrM location, BrM size, or grade of BrM (Table 3). In subsequent analyses, cases with AR expression ≥ 10% were labelled as "AR positive (AR+)", while those with AR expression < 10% were labelled as "AR negative (AR−)". This was based on a previous study that found that AR expression threshold of 10% is associated with clinical response to enzalutamide in TNBC [8]. 56% (n = 32/57) of BrM were AR positive. Distribution of BrM subtype, Ki-67 expression, GATA3 expression, age at BrM diagnosis, BrM location, BrM size, and BrM grade, as stratified by AR status, is shown in Table 3. No significant associations were found between AR status and age at BrM diagnosis, BrM location, BrM size, or grade of BrM (Table 3).

AR Expression in Relation to Ki-67 Expression and GATA3 Expression
We found a significant association between Ki-67 expression by IHC and both frequency of AR expression (Figure 2a, p = 0.020) and mean absolute AR expression (Figure 2b, p = 0.005). The subgroup with high (50% or above) Ki-67 expression had the lowest AR expression (p = 0.005 for Tukey HSD between Ki-67 high vs. Ki-67 intermediate groups) (Figure 2b).
We also found a significant association between GATA3 expression by IHC and both frequency of AR expression (Figure 2c, p = 0.003) and mean absolute AR expression (Figure 2d, p = 0.005). The subgroup with high (50% or above) GATA3 expression had the highest AR expression (p = 0.017 for Tukey HSD between GATA3 high vs. GATA3 negative groups, p = 0.035 for Tukey HSD between GATA3 high vs. GATA3 low groups) (Figure 2d).

AR Expression in Matched Primary Breast Cancers
In 10 patients for whom matched BrM and primary breast cancers were available, AR expression by IHC was higher in primary breast cancers compared to matched BrM (79% vs. 52%, Wilcoxon signed rank test p < 0.05). 90% (n = 9/10) of primary breast cancers were AR+, whereas 70% (n = 7/10) of BrM were AR+. AR status was concordant in 7 of 10 cases. Among the 3 cases for which AR status differed in the primary cancer and BrM, AR was "gained" in the BrM in one case and "lost" in the BrM in two cases (Figure 3). We also found a significant association between GATA3 expression by IHC and both frequency of AR expression (Figure 2c, p = 0.003) and mean absolute AR expression (Figure 2d, p = 0.005). The subgroup with high (50% or above) GATA3 expression had the highest AR expression (p = 0.017 for Tukey HSD between GATA3 high vs. GATA3 negative groups, p = 0.035 for Tukey HSD between GATA3 high vs. GATA3 low groups) (Figure 2d).

AR Expression in Matched Primary Breast Cancers
In 10 patients for whom matched BrM and primary breast cancers were available, AR expression by IHC was higher in primary breast cancers compared to matched BrM (79% vs. 52%, Wilcoxon signed rank test p < 0.05). 90% (n = 9/10) of primary breast cancers were AR+, whereas 70% (n = 7/10) of BrM were AR+. AR status was concordant in 7 of 10 cases. Among the 3 cases for which AR status differed in the primary cancer and BrM, AR was "gained" in the BrM in one case and "lost" in the BrM in two cases (Figure 3).

Associations between AR Expression and Clinical Outcomes
Patients with AR+ BrM had a median OS of 12.5 months, compared to patients with AR− BrM who had a median OS of 7.9 months (p = 0.6) (Figure 4a). Patients with AR+ BrM had a median bsPFS of 8.0 months, compared to patients with AR− BrM who had a median bsPFS of 5.1 months (p = 0.95) (Figure 4b). These differences were not statistically significant.

Associations between AR Expression and Clinical Outcomes
Patients with AR+ BrM had a median OS of 12.5 months, compared to patients with AR− BrM who had a median OS of 7.9 months (p = 0.6) (Figure 4a). Patients with AR+ BrM had a median bsPFS of 8.0 months, compared to patients with AR− BrM who had a median bsPFS of 5.1 months (p = 0.95) (Figure 4b). These differences were not statistically significant.

Discussion
AR expression in breast cancer BrM is of interest due to readily available androgentargeted therapies that can cross the blood brain barrier. This study demonstrated in a retrospective Canadian cohort that AR is expressed by IHC in the majority of breast cancer BrM, and that HER2+ BrM most frequently expressed AR, while TNBC BrM least frequently expressed AR. AR status was concordant between the majority of BrM and matched primary breast tumours. AR status of the BrM was not associated with OS or bsPFS, although patients with AR+ BrM had a trend towards developing BrM later in the course of their disease.
In our study, AR expression by IHC was expressed (i.e., expression by IHC > 0% using SP107) in 64.9% of breast cancer BrM. This is higher than what was reported in a previous study which demonstrated that AR expression was expressed (i.e., expression by IHC > 0% using AR441) in 35.1% of breast cancer BrM in a similar-sized cohort [17]. The antibody used in our study (SP107) has been shown to have higher sensitivity and robustness compared to AR441 [8]. In our analyses, we also used a clinically relevant AR threshold of ≥10% expression, which corresponds to the threshold associated with clinical Patients with AR+ BrM had a numerically longer median time from diagnosis of breast cancer to diagnosis of BrM, compared to those with AR− BrM, although this was not statistically significant (51 vs. 29 months, p = 0.16) (Figure 4c).

Discussion
AR expression in breast cancer BrM is of interest due to readily available androgentargeted therapies that can cross the blood brain barrier. This study demonstrated in a retrospective Canadian cohort that AR is expressed by IHC in the majority of breast cancer BrM, and that HER2+ BrM most frequently expressed AR, while TNBC BrM least frequently expressed AR. AR status was concordant between the majority of BrM and matched primary breast tumours. AR status of the BrM was not associated with OS or bsPFS, although patients with AR+ BrM had a trend towards developing BrM later in the course of their disease.
In our study, AR expression by IHC was expressed (i.e., expression by IHC > 0% using SP107) in 64.9% of breast cancer BrM. This is higher than what was reported in a previous study which demonstrated that AR expression was expressed (i.e., expression by IHC > 0% using AR441) in 35.1% of breast cancer BrM in a similar-sized cohort [17]. The antibody used in our study (SP107) has been shown to have higher sensitivity and robustness compared to AR441 [8]. In our analyses, we also used a clinically relevant AR threshold of ≥10% expression, which corresponds to the threshold associated with clinical response to enzalutamide in TNBC [8].
With respect to AR expression by IHC in primary breast cancers, our results are comparable to those of other studies, which consistently report that AR is expressed in 72-79% of primary breast cancers [3][4][5]. In our analyses of BrM, AR is most frequently expressed in HER2+ BrM (previously reported to be expressed in 76-87% of primary HER2+ breast cancers [3,19,20]), and AR is least frequently expressed in TNBC BrM (previously reported to be expressed in 27-35% of primary TNBC) [3,5,7]. The largest among these studies, which was performed on 2171 patients in the US between 1976 and 1996 using the AR441 antibody, reported AR expression (≥1%) in 91% of Luminal A, 68% of Luminal B, 59% of HER2-postive, and 32% of triple negative primary breast cancers [5]. We observed a similar pattern of AR gain and loss between primary breast cancers and matched metastases as a previous study, which showed that AR was expressed ≥10% by SP107 in 82.9% of 164 primary breast tumours and 60.2% of 83 metastases, with a concordance rate of 60.6% between primary breast cancers and matched metastases [21]. The variability in concordance between the AR status of primary breast cancers and matched metastases suggests that re-checking AR status in metastatic sites may be of value when considering the use of an AR-targeted therapy.
While AR expression by IHC is the gold standard, previous studies have also investigated mRNA expression of AR in primary breast cancer, although to our knowledge there have been no such studies in breast cancer BrM. A study of 101 primary TNBCs found that qRT-PCR detected AR expression in 34% of cases compared to 15% of cases using IHC (i.e., ≥1% using AR441 antibody), with 75% concordance between the two methods; this study found that AR mRNA expression was associated with shorter distant metastases free survival, which is consistent with previous studies employing IHC in TNBC [22]. Another study of 872 primary ER-negative breast cancers found that mRNA expression of AR had a strong correlation (R = 0.68, p < 0.001) with AR reverse phase protein assay, and AR mRNA expression was associated with lower grade disease and lower risk of recurrence [23]. Interestingly, this study also showed that AR mRNA expression was associated with HER2+ status, which is consistent with our finding that HER2+ BrM have highest expression of AR by IHC. While qRT-PCR may be a more sensitive method that can identify patients whose tumours have low levels of AR expression not detected by IHC, the likelihood that such patients would benefit from androgen-directed therapies is uncertain, given that AR IHC expression thresholds for clinical response to androgen-directed therapies are relatively high, e.g., ≥10% for clinical response to enzalutamide in TNBC [8], and ≥40% for clinical response to enobosarm in HR+ breast cancer [24]. Presently, IHC remains the most extensively studied and most clinically relevant method for assessing AR expression.
Our study found no significant association between AR expression with either OS or bsPFS, which is not surprising given that our cohort consisted of multiple different breast cancer subtypes, which inherently portend different prognoses; the number of patients within each subtype was too small to draw meaningful conclusions. Additionally, no patients in our study received AR-targeted therapies.
AR signaling plays a different biological role in different breast cancer types [11]. In ER+ breast cancer, previous studies have shown that positive AR status was associated with longer breast-specific survival [25,26]. This is not surprising given that AR functions as a tumour suppressor in ER+ breast cancer. The activation of AR represses ER-regulated cell cycle genes and upregulates AR target genes which include tumour suppressors [27], and this supports the rationale for ongoing trials evaluating AR agonists in ER+ breast cancer (Table 3). In TNBC, the prognostic significance of AR is more controversial [28][29][30][31][32]. In TNBC, AR binds to androgen-related element in the nucleus to induce cell proliferation [33,34]. This supports the role for ongoing trials evaluating AR antagonists in TNBC (Table 3), with the caveat that the utilization of robust biomarkers for optimal patient selection is paramount. For example, luminal AR+ TNBCs express high levels of AR and are particularly sensitive to bicalutamide in preclinical models [35]. Finally, in HER2+ breast cancer, enzalutamide inhibited growth in trastuzumab-resistant HER2+ breast cancer xenografts [36], which may be attributed to the cross-talk between AR signaling and HER2 signaling [34]. A recent phase II clinical trial demonstrated that enzalutamide plus trastuzumab was well tolerated in patients with AR+/HER2+ breast cancer who were pretreated with anti-HER2 therapy, and a subset of these patients had durable disease control [13]. However the median PFS was only 3.4 months and patients with BrM were not included in this trial [13]. Our study showed that HER2+ breast cancer BrM most frequently express AR, and suggests that there is a need to elucidate the biology of AR in HER2+ breast cancer.
Many of the associations that we found between the expression of AR and the expression of other prognostic and proliferation markers (GATA3 and Ki-67) are in keeping with what is known about the biology of breast carcinomas. Ki-67 is a nuclear protein associated with cellular proliferation, and is associated with higher histological grade and shorter survival [37]. The fact that we did not find a linear relationship between Ki-67 and AR expression is not surprising, given that AR plays opposing roles in proliferation among different breast cancer subtypes. The subgroup with high (50% or above) Ki-67 expression had the lowest AR expression, and the majority (5 of 9) of BrM which were Ki-67 high/AR low were TNBC. This is consistent with the fact that TNBC has a higher rate of proliferation and less frequent expression of AR. GATA3 is a transcription factor that plays a role in the differentiation of breast luminal epithelial cells [38], and is associated with ER positive status as well as favorable prognosis [39,40]. In our study, the subgroup with high (50% or above) GATA3 expression had the highest AR expression; not surprisingly, 17 of 19 BrM which were GATA3 high/AR positive were also HR+.
The fact that our study shows a high proportion of breast cancer BrM are AR-positive using a biomarker predictive of response to anti-androgen therapies, supports the evaluation of AR-targeted therapies in breast cancer patients BrM. Table 3 summarizes 16 currently active clinical trials investigating AR-targeted therapies in advanced and metastatic breast cancer. It is encouraging that the majority (eleven of sixteen) of these trials include patients with controlled BrM. Six trials evaluate androgen agonist therapies (enobosarm, EP0062, EG017 and CR 1447); out of these trials, the majority (five of six trials) include patients with ER+/HER2− breast cancer, and two include patients with TNBC. Ten trials evaluate androgen antagonist therapies (enzalutamide, bicalutamide, and seviteronel); out of these trials, the majority (seven of ten trials) include patients with TNBC, five include patients with ER+/HER2− breast cancer, and only one includes patients with HER2+ breast cancer (and this trial excluded patients with BrM). Given that previous trials have shown the clinical response rates of antiandrogens in AR+ TNBC to be underwhelming, in the range of 19-35% [8][9][10][11][12], it is important that ongoing and future trials evaluate potential predictive biomarkers of response to optimize patient selection, as well as investigate combination therapies.
Our study is limited by the small retrospective single centre design, and the fact that no anti-androgen therapies were prescribed to our patients. The small number of patients with matched primary and BrM tissue is another limitation. In the future, pre-clinical models of AR-positive BrM would be useful in validating these findings.

Conclusions
This study used a clinically validated antibody (SP107) to show in a 57-patient retrospective Canadian cohort that the majority of breast cancer BrM were AR positive (defined as AR ≥ 10% by IHC). Furthermore, AR expression was most frequently expressed in HER2+ BrM and least frequently expressed in TNBC BrM. These findings are comparable to AR expression previously reported in primary breast cancers. HER2+ BrM most frequently express AR, yet there are no active clinical trials evaluating AR-directed therapies in patients with HER2+ BrM; hence, additional trials for this population are warranted. While AR-targeted therapies may be most clinically useful in the treatment of TNBC, AR is expressed least frequently in TNBC BrM, suggesting a need for improved biomarkers to judiciously guide patient selection. Overall, our study shows that AR is a frequently expressed and promising intracranial molecular target, supporting the investigation of androgen-targeted therapies in the treatment of patients with breast cancer BrM. Informed Consent Statement: Patient consent was waived due to the retrospective nature of the study without patient identifying information and because all participants in the study were deceased.

Data Availability Statement:
The data presented in this study are available in this article.