The impact of androgen receptor (AR) and histone deacetylase 1 (HDAC1) expression on the prognosis of ductal carcinoma in situ (DCIS)

Background Ductal carcinoma in situ (DCIS) display favorable outcome but little is known about the factors associated with invasive recurrence. To identify better prognostic biomarkers, we performed gene expression analysis followed by immunohistochemistry (IHC) staining validation. Methods Differential gene expression (DGE) analysis of 24 pure DCIS patients was performed using a nanostring platform. RNA was extracted from paran blocks from age/estrogen receptor matched recurrence-free (n=16) and invasive-recurrence (n=8) cases (disease-free interval >5 years). External validation was done among independent 61 cases, invasive-recurrence (n=16) and recurrence-free (n=45) pure DCIS cases by IHC staining. Results Eight differentially expressed genes were found statistically signicant (log 2-fold change <–1 or >1 and p<0.001). Less than ½ fold expression of CUL1, AR, RPS27A, CTNNB1, MAP3K1, PRKACA, GNG12, MGMT genes were observed in REC cases compared to NED cases. Androgen receptor (AR) and histone deacetylase 1 (HDAC1) were selected for external validation (AR: log 2-fold change –1.35, p<0.001, and HDAC1; log 2-fold change –0.774, p<0.001). AR and HDAC1 protein expression was externally validated by IHC staining of 61 pure DCIS cases (16 invasive-recurrence versus 45 recurrence-free). Absence of AR and high HDAC1 expression was an independent risk factor for invasive recurrence (hazard ratio 5.04, 95% CI: 1.24, 20.4; p=0.023, hazard ratio 3.07, 95% CI: 1.04, 9.04; p=0.042). High nuclear grade (NG 3) was also associated with long term invasive recurrence. of HDAC1 was associated with greater risk of long term invasive recurrence among pure DCIS. AR


Introduction
Ductal carcinoma in situ (DCIS) consist upto 20% of newly diagnosed breast cancers owing to widely adopted screening. It is a non-obligate precursor to invasive ductal carcinoma (IDC) exhibiting an excellent prognosis. While surgical resection is the mainstay of therapy, adjuvant radiotherapy (RT) after lumpectomy has long been a controversial issue. Trials showed that the relative reduction in local recurrence rate was 50% in the irradiated group.
But, these studies failed to show bene t on distant metastases or overall survival [1]. RT may induce cardiac toxicity, brosis of lung and skin changes. It needs daily visits to the radiation center and also costs substantial medical expenses. The risk could outweigh the bene t in patients with a low risk implying the necessity of accurately classify patient groups according to risk of subsequent invasive recurrences. The absolute bene t of chemoprevention is being investigated as well taking into account the side effects of uterine carcinoma and menopausal symptoms. Yet, owing to lack of tool to accurately classify risk and bene t of such treatment, current guideline recommends surgical resection followed by radiation and/or chemoprevention with tamoxifen albeit the indolent clinical course of the disease.
Young age, high histologic grade, large tumor size and absence of hormone receptor are the historical prognostic markers of DCIS [2][3][4]. Van Nuys prognostic index presents the risk of invasive recurrence into low/intermediate/high risk group by adding scores of patient age, tumor size, margin width and nuclear grade/necrosis [5]. Oncotype DCIS predicts prognosis using gene expression of DCIS tumor samples [6]. DCISionRT is a recently developed radio-genomic tool which predicts the bene t of radiation [7]. However, both requires higher level of evidence and overall feasibility including the cost-effectiveness [8]. For DCIS, it is important to assess the risk of long term recurrence as upto 98% display disease-free within ve year after diagnosis. In this study, we aimed to investigate prognostic biomarkers of DCIS to predicting long term invasive recurrence. We sought to identify markers which could be easily applied in clinic using immunohistochemistry.

Patient Selection
We performed a retrospective study with mRNA expression pro ling and external validation by IHC staining. A discovery cohort was selected for gene expression analysis and a validation cohort was subsequently chosen for IHC staining. Patients histologically diagnosed with pure DCIS without an invasive component after de nitive surgery were selected. As to investigate the long term prognosis, all patients underwent surgery before 2008 were analyzed. Patients with clinicopathological data and formalin xed para n embedded available were included.
Thirty-six pure DCIS patients were selected for discovery cohort. Twelve patients with invasive recurrence (REC) were chosen. Cases with no recurrence (NED) were selected by 1:2 ratio matching for age and ER. All patients were diagnosed with DCIS and underwent de nitive surgery with negative tumor margin after breast-conserving surgery or mastectomy, between 1995 and 2004. Sequencing was successfully performed in 8 REC cases and 16 NED cases. Sixty-one pure DCIS patients were selected as validation cohort for IHC analysis. REC and NED cases were selected as done for discovery cohort. All patients have underwent surgery at Asan Medical Center (AMC) between 1995 and 2008.

RNA Extraction And Quanti cation
Surgical specimens from the discovery cohort obtained during surgery for routine diagnostic pathologic examinations were used. Areas of tumor component were dissected from formalin-xed para n-embedded (FFPE) tissue. Total RNA was extracted from the sections using MasterPure™ Complete DNA & RNA Puri cation Kit (Lucigen-Epicentre, Middleton WI, USA), according to the manufacturer's instructions. RNA yield and purity were assessed using a DS 11 Spectrophotometer (Denovix Inc., Wilmington DE, USA). Since nanostring recommends the use of ≈ 300 ng total RNA, 300 ng of total RNA was added to the sample preparation reaction in the available 5 µL volume. An RNA quality check was performed by using Fragment Analyzer (Advanced Analytical Technologies, Ankeny, IA, USA). mRNA Expression pro ling by nanoString nCounter system and data analysis The digital multiplexed nanoString nCounter human mRNA expression assay (nanoString Technologies, Seattle, WA, USA) was performed with 300 ng total RNA isolated from the tissues. Samples were assessed using the nCounter PanCancer Pathways Panel, which targets 730 genes representing all the major cancer pathways. Hybridizations were carried out by combining 5 µL of each RNA sample with 8 µL of nCounter Reporter probes in hybridization buffer and 2 µL of nCounter Capture probes (for a total reaction volume of 15 µL) overnight at 65℃ for 16-20 h. Excess probes were removed using two-step magnetic bead-based puri cation on the nCounter Prep Station (nanoString Technologies).
Abundances of speci c target molecules were quanti ed on the nCounter Digital Analyzer by counting the individual uorescent barcodes and assessing the target molecules. For each assay, a high-density scan encompassing 280 elds of view was performed. Data were collected using the nCounter Digital Analyzer after taking images of the immobilized uorescent reporters in the sample cartridge with a charged-couple device camera. mRNA data analysis was performed using the nSolver software analysis, freely available from nanoString Technologies. The mRNA pro ling data were normalized using housekeeping genes. R software (R Core Team.

Results
Differentially expressed genes among the discovery cohort Among the initially selected 36 pure DCIS patients for comparative mRNA expression pro ling, sequencing was successfully performed in 8 REC cases and 16 NED cases. Clinicopathological characteristics of the 24 cases (8 REC, 16 NED) are displayed in Table 1. The characteristics of the REC and NED group showed no statistically signi cant differences. Among the REC patients, 6 cases were local recurrence and 2 were regional recurrence. The median follow-up was 149 months in total: 56 months for the REC group, and 162 months for the NED group. The mRNA expression levels for these two groups were compared. Fold change (FC) was calculated as the ratio of gene expression level REC/NED. Genes with log2_|FC| >1 and p value < 0.001 were selected as differentially expressed genes (DEG). Recurred cases had signi cantly lower expression of CUL1, AR, RPS27A, CTNNB1, MAP3K1, PRKACA,  Cutoff score: Log2-fold change <-1 or > 1 and p < 0.001.
Among the DEGs, AR, gene coding androgen receptor was chosen as a candidate prognostic biomarker. Androgen receptor (AR) is a hormone receptor that is expressed in over 70% of invasive breast cancers [9]. While it's role during breast cancer carcinogenesis is not clear, AR is suggested to have an inhibitory effect especially in estrogen receptor (ER)-positive breast cancer. It is an emerging prognostic and predictive marker and a potential therapeutic target for invasive breast cancers [10]. Gene expression data of the discovery cohort presented 1.71 times over expression of HDAC1 gene, Histone deacetylase 1 in recurred cases compared to NED group (log2_FC 0.774, p < 0.001). HDAC1 is a member of the protein family of HDAC, which is involved in chromatin modi cation by removing acetyl groups from histones. It is known to contribute to cancer cell development by suppressing the expression of genes such as cell cycle inhibitors, tumor suppressor genes, and apoptosis inducer genes, and by inducing the global loss of acetylation [11]. HDAC1 inhibitors induced cell differentiation and inhibited the growth of breast cancer cells by cell cycle arrest and induction of apoptosis [12]. We evaluated the prognostic impact of both AR and HDAC1 within an independent validation cohort by IHC.

Validation of prognostic biomarker among an independent cases
An independent cohort was selected for validation of AR and HDAC1 as a prognostic biomarker. Samples from 61 pure DCIS surgical specimens were obtained for IHC staining. Among the 61 validation cohort, 16 were REC patients and 45 were NED patients. Clinicopathological characteristics of the validation cohort is described in Table 3. In the REC group, 13 local recurrence cases and 3 regional lymph node recurrence patients were included. The median follow-up time was 110 months in total: 65 months for the REC group, and 114 months for the NED group. Figure 1 shows the IHC staining results for AR. To evaluate factors associated with AR negativity (versus AR positive) and high HDAC1 expression (versus HDAC1 low) the validation cohort was divided according to AR-positive versus ARnegative group, and a high HDAC1 versus low HDAC1 group. As indicated in Table 4, each pair of groups were statistically homogeneous regarding other conventional risk factors.   Survival analysis was done to evaluate AR and HDAC1 protein expression and the risk of invasive recurrence. Figure 2 illustrates Kaplan-Meier survival curve for AR negative versus positive group, and HDAC1 high and low group. 10 year invasive recurrence for AR negative versus positive, and HDAC1 high versus low was 53.3% vs 77.6% and 46.8% vs 83.5%. Factors associated with recurrence by univariate and multivariate Cox regression analysis is described in Table 5. Conventional risk factors eg, age and size, were not signi cantly associated. Patients with NG 3 had a higher recurrence rate than NG 1or 2 cases indicating high grade DCIS as an independent risk factor (HR 4.89, p = 0.005). Both AR-negativity and high HDAC1 expression were independently associated with invasive recurrence (HR 5.04, p = 0.031, and HR 3.07, p = 0.010, respectively). We then expanded the survival analysis including the IHC result of the discovery cohort. Similarly, both AR negativity and high expression of HDAC1 by IHC was an independent risk factors (p = 0.004, and p = 0.038, data not shown)

Discussion
To identify biomarkers for long term invasive recurrence in pure DCIS, we performed differential mRNA expression pro ling of cancer panel genes followed by external validation using IHC for selected candidates, AR and HDAC1.
The study was conducted with pure DCIS cases and showed that DCIS cases that progress into invasive carcinoma present different gene expression features compared to DCIS cases that remain in the NED state. Differential gene expression pro ling found 8 genes that were found to be less expressed in REC cases (|FC|>2, p < 0.001). Among them, AR and HDAC1 was selected for external validation by IHC, the most commonly applied methodology in realworld clinic. Survival analysis revealed that the group with the absence of AR cells and group with higher HDAC1 exhibited higher risk of invasive recurrence. These results were statistically signi cant, suggesting that both AR and HDAC1 may serve as potential prognostic biomarkers for pure DCIS patients.
Unlike ER, the role of AR in the carcinogenesis of breast cancer is not well established. AR is co-expressed with ER in 70-90% of cases and is expressed in 40% of ER-negative cancers. The effect of AR in ER-negative breast cancer is rather mixed, but in ER-positive invasive breast cancers, cross talk between ER signaling and AR signaling do exist. Several studies showed that AR expression is associated with favorable characteristics such as, older age, smaller size, well-differentiated tumors, lower proliferation index, and higher positivity of hormone receptors and as a consequence, better overall survival [13][14][15][16]. Results from in vitro models support the ndings that overexpression of AR in ER-positive cell lines inhibits the proliferative activity of ER [17]. AR appears to have a competitive effect against ER for binding to the estrogen-response element of ER target genes [18].
However the prognostic role of AR in DCIS is more ambiguous. IHC staining of hormone receptors showed that the rate of AR expression was lower in high-grade (HG)-IDC than in HG-DCIS [19]. Other studies showed that AR expression level was higher in DCIS adjacent to IDC compared to pure DCIS [20,21]. In statistical analyses of the expression of several biomarkers of breast cancer, AR did not show statistical signi cance as a risk factor for recurrence [22].Recently, based on the concept of competition between AR and ER, Ravaioli et al. suggested AR/ER ratio as a prognostic marker for DCIS [23]. AR indeed was associated with a favorable prognosis in this study population which correlates with the previously reported biological effect of AR as an ER signal inhibitor. As the cases were selected according to HR positivity rather than in a consecutive manner, we couldn't directly see the effect of AR con ned to HR positive DCIS cases. Also the numbers analyzed were small especially the REC cases.
The HDAC family is an epigenetic regulator of gene expression that plays an important role in both normal cell development and carcinogenesis. In cancer cells, by removing an acetyl group from histones, HDAC causes chromatin condensation resulting in the global gene expression suppression including tumor suppressor genes, cell cycle inhibitors, and differentiation factors or apoptosis inducers. HDAC1 regulates the expression of key proteins involved in the cell cycle, such as p21, p53, and cyclin D1. Also, overexpression of HDAC1 induces hypoxia-inducible factor-1α and vascular endothelial growth factor and increases angiogenesis. Besides speci c gene regulation, global loss of acetylation at Lys16 and trimethylation at Lys20 of histone H4 are hallmarks of human cancer cells, including breast cancer cells [24]. Moreover, global loss of histone modi cation was related to poor breast cancer-speci c survival and disease-free survival [25]. Little is known about the correlation between HDAC1 and breast cancer. In an in vitro study, overexpression of HDAC1 induced the loss of ER and increased cell proliferation, and the use of an HDAC inhibitor induced ER re-expression [26]. Junko Suzuki reported DCIS showing reduced acetylation compared to normal breast epithelial cells, regardless of HDAC1 expression [27]. Several retrospective studies showed HDAC1 to be associated with a favorable prognosis [28]. A recent meta-analysis argued that HDAC1 overexpression did not correlate with disease-free survival and overall survival in invasive breast cancer [29]. The expression of HDAC1 in DCIS remains controversial.
This study implies the potential role of AR and HDAC1 as prognostic biomarkers for pure DCIS. The strength of this study is that the cohort consist of pure DCIS cases with long term follow up, as most of DCIS patients don't undergo > 5-10 year surveillance. Observation of pure DCIS lesions over a long follow-up period, compared to analyzing synchronous DCIS lesions, enables investigating the natural history more directly. Synchronous DCIS lesions may not be a high-risk DCIS and may possess different biological features. Also, to represent true high risk group, in situ recurrences were excluded and only invasive recurrence were included for REC group. Nonetheless, there are several limitations. The number of cases analyzed in both cohorts were small. The study was not a consecutive case-series analysis. Rather, each of REC and NED cases were selected according to age, hormone receptor status in a 1:2 ratio. But as DCIS is a disease displaying exceptionally favorable outcome, cases with invasive recurrence within certain timeframe is small. The REC and NED cases were selected matched by age, hormone receptor status as to minimize the effect of these known risk factors. The cutoff value of AR and HDAC1 should be furtherly addressed as, currently there is no standard cutoff.

Conclusion
In conclusion, we have performed a matched comparative gene expression analysis of pure DCIS cases to identify prognostic indicators of long term invasive recurrence. Our study revealed absence of AR and overexpression of HDAC1 was found to be associated with greater risk of invasive recurrence. Further validation within a larger series are needed.

Declarations
Ethics approval and consent to participate Clinicopathological information and follow up recurrence data was reviewed within ABLE (Asan Biomedical Research Environment) system which enable anonymous data and sample extraction. The study was approved by the institutional review board (IRB No. 2016-0976) of AMC.

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Kaplan-Meier survival analysis. Recurrence-free survival curve. (a) Group with AR negative showed poor prognosis (p-value 0.031, 10yr DFS 53.3%, 77.6% for AR negative and positive group respectively ).(b) Group with high HDAC1 showed higher recurrence rate ( p -value 0.010, 10yr DFS 46.8%, 83.5% for HDAC1 high and low group respectively).

Competing interests
DFS Disease free survival, AR androgen receptor, HDAC1 histone deacetylase 1