Predicting Radiation Resistance in Breast Cancer with Expression Status of Phosphorylated S6K1

Emerging evidence suggests that the mammalian target of rapamcyin (mTOR) pathway is associated with radio-resistance in cancer treatment. We hypothesised that phosphorylated ribosomal S6 kinase 1 (p-S6K1), a major downstream regulator of the mTOR pathway, may play a role in predicting radio-resistance. Therefore, we evaluated the association of p-S6K1 expression with radio-resistance in breast cancer cell lines and patients. During median follow-up of 33 (range, 0.1–111) months for 1770 primary breast cancer patients who underwent surgery, patients expressing p-S6K1 showed worse 10-year loco-regional recurrence-free survival (LRFS) compared to that of p-S6K1-negative patients after radiotherapy (93.4% vs. 97.7%, p = 0.015). Multivariate analysis revealed p-S6K1 expression as a predictor of radio-resistance (hazard ratio 7.9, 95% confidence interval 1.1–58.5, p = 0.04). In vitro, CD44high/CD24low MCF7 cells with a radioresistant phenotype expressed higher levels of p-S6K1 than control MCF7 cells. Furthermore, the combination of radiation with treatment of everolimus, an mTOR-S6K1 pathway inhibitor, sensitised CD44high/CD24low MCF7 cells to a greater extent than MCF7 cells. This study provides in vivo and in vitro evidence for p-S6K1 expression status as an important marker for predicting the resistance to radiotherapy and as a possible target for radio-sensitization in breast cancer patients.

During the last few decades, advances in the molecular profiling of breast cancer have substantially changed treatment approaches towards personalised therapy, and provided patients with significantly improved survival rates and quality of life 1 . Despite these enormous strides in tailoring a variety of approaches to systemic therapies, the role of radiotherapy in precision medicine for breast cancer is only beginning to emerge [2][3][4] . Technical advances in radiotherapy have resulted in an improved ability to decrease dose delivery to surrounding vital organs such as the heart and lungs while accurately targeting the diseased tissue. In addition, a number of dose fractionation strategies have now been validated for patients with early-and advanced-stage breast cancer. Partial breast irradiation via external beam, brachytherapy, or intraoperative radiation techniques has been shown to limit the volume of irradiated tissue in selected patients while preserving efficacy [5][6][7] . However, in terms of biology, uniform doses of radiation have been typically delivered without consideration to differences across breast cancer subtypes. Consequently, the majority of breast cancer patients receive radiotherapy according to the type of surgery and clinical stage rather than based on their sensitivity to this kind of therapy.
Radiotherapy is a critical treatment modality for achieving loco-regional control in breast cancer patients. Moreover, the pivotal role of adjuvant radiotherapy in leading to reduced recurrence and long-term mortality has been well-established through the analyses of the Early Breast Cancer Trialists' Collaborative Group 8,9 . Nevertheless, some patients still may not benefit from this treatment owing to individual variation in radio-sensitivity and may experience recurrences that ultimately challenge their prognosis and quality of life. It is therefore necessary to develop new biomarkers that predict the effectiveness of radiotherapy. Prediction of radio-resistance in individual patients may enable optimised dose modifications or application of radio-sensitisers www.nature.com/scientificreports www.nature.com/scientificreports/ to tumours with radio-resistance 2 . Although several tumour-related descriptive factors such as vascular invasion, larger tumour size, and higher grade have been considered as predictors of loco-regional recurrence to date, they provide limited information about the intrinsic tumour response to radiotherapy 10 .
Phosphorylated ribosomal S6 kinase 1 (p-S6K1) is a key downstream effector of the mammalian target of rapamycin (mTOR) pathway and has been considered an indirect marker of mTOR activity [11][12][13] . Because p-S6K1 relays many signals to promote further oncogenic translations in tumourigenesis 14 , upregulation of p-S6K1 in the tumour has been suggested to be predictive of resistance to systemic therapies such as endocrine therapy and chemotherapy in breast cancer patients [15][16][17] . However, few studies on p-S6K1 have focused on its role in the response to radiotherapy [18][19][20] . As mounting evidence indicates that the mTOR pathway is associated with the development radio-resistance in tumours [21][22][23] , as an indicator of mTOR activity p-S6K1 might be related with the development of radio-resistance in breast cancer.
Therefore, in this study, we investigated the association between the expression of p-S6K1 and radio-resistance in breast cancer patients and evaluated the clinical potential of p-S6K1 in predicting radio-resistance. Furthermore, we explored the association between p-S6K1 and radio-resistance in breast cancer stem cells.

Results
Patient characteristics. The median age of the patients included in this study was 51 years. The majority of patients who received radiotherapy (n = 1317, 74.4%) had breast conserving surgery (n = 1044, 79.3%), whereas the remainder underwent total mastectomy (Table 1). Positive and negative p-S6K1 expression was detected in 1323 (74.4%) and 447 (25.3%) patients, respectively. The p-S6K1 expression status was evenly distributed among patients treated with or without radiotherapy (p = 0.091). The numbers of loco-regional recurrences were also well distributed between the two groups (p = 0.094).
Association of p-S6K1 expression status and loco-regional recurrence. Patients with positive p-S6K1 expression showed worse loco-regional recurrence-free survival (LRFS) compared to those with negative p-S6K1 expression when treated with radiotherapy. The 10-year LRFS rate in patients treated with radiotherapy was 93.4% for pS6K1-positive patients and was 97.7% for pS6K1-negative patients (p = 0.015, Fig. 1a). However,  no association between p-S6K1 expression status and LRFS was observed when patients did not receive radiotherapy (p = 0.702, Fig. 1b). Loco-regional control rates at 10 years were 94.6% and 89.0% in patients with and without radiotherapy, respectively. In univariate analysis to identify predictors of radio-resistance, positive p-S6K1 expression was found to be predictive of worse LRFS (p = 0.024) in patients treated with radiotherapy. After multivariate analysis, positive p-S6K1 expression (p = 0.044) remained a predictor of worse LRFS (Table 2). However, as mass size (p = 0.038) was also predictive of worse LRFS in the multivariate analysis, we cross-analysed mass size and p-S6K1 expression status. Ten loco-regional recurrences occurred in 779 patients with tumours ≤2 cm, and 28 recurrences occurred in 538 patients with tumours >2 cm (Supplementary Table S1). When the tumours were >2 cm, significantly worse LRFS was observed in patients with a positive p-S6K1 status compared to that of patients with a negative p-S6K1 status (p = 0.044). In contrast, there was no significant difference in LRFS according to p-S6K1 status in patients with tumours ≤2 cm (p = 0.202, Supplementary Fig. S1). However, among the 10 recurrences that occurred in tumours ≤2 cm, nine (90%) tumours were p-S6K1-positive, whereas only one (10%) tumour had a negative p-S6K1 status. Figure 1. Loco-regional recurrence-free survival (LRFS) according to p-S6K1 expression status in patients with or without radiotherapy. Kaplan-Meier estimates of LRFS among patients treated with (a) and without (b) radiotherapy. Patients with positive p-S6K1 expression showed significantly worse LRFS compared to those with negative p-S6K1 expression when treated with radiotherapy (*p = 0.015). In contrast, no difference in LRFS based on p-S6K1 expression status was seen in patients who did not receive radiotherapy (*p = 0.702). *Kaplan-Meier survival estimate compared by a log-rank test. www.nature.com/scientificreports www.nature.com/scientificreports/ Increased p-S6K1 could be associated with radio-resistance in breast cancer stem cells. Because our clinical analysis showed a positive correlation between p-S6K1 expression and radio-resistance in breast cancer patients, we further examined whether the S6K1 pathway regulates the radio-resistance of breast cancer in vitro. First, we determined the expression levels of phosphorylated-mTOR (p-mTOR), a direct upstream regulator of S6K1, and p-S6K1 in normal breast cells and several breast cancer cell lines. We found that the phosphorylated and endogenous levels of mTOR and S6K1 were generally increased in breast cancer cells compared to those of normal breast MCF10A cells (Fig. 2a). Of the breast cancer cell lines evaluated, MCF7 cells expressed higher levels of phosphorylated and endogenous mTOR and S6K1 (Fig. 2a). Furthermore, we found that CD44 high /CD24 low MCF7 cells, a breast cancer stem cell line with a radioresistant phenotype 24,25 (Supplementary Fig. S2), expressed higher levels of p-S6K1 compared to those of control MCF7 cells (Fig. 2b), suggesting that high expression levels of p-S6K1 are involved in the radio-resistance of breast cancer. Thus, control and CD44 high /CD24 low MCF7 cells were used to further examine whether the mTOR-S6K1 pathway regulates the radio-resistance of breast cancer. Notably, we found that combining radiation with treatment of everolimus, a clinically available inhibitor of the mTOR-S6K1 pathway 26 , sensitised the radio-resistant CD44 high /CD24 low MCF7 cells to a greater degree than observed in the control MCF7 cells (Fig. 2b,c). Therefore, our data suggest that increased levels of p-S6K1 expression could be associated with radio-resistance in breast cancer stem cells.

Discussion
In this study, we provide in vitro and clinical evidence that p-S6K1 expression status could be associated with a reduced response to radiotherapy in breast cancer. To the best of our knowledge, our study is the first to demonstrate the potential of p-S6K1 expression status as a marker for radio-resistance in breast cancer. To date, only a few studies have provided indirect evidence that S6 kinase may be associated with the response to radiotherapy. However, our analysis directly compared the outcomes of radiotherapy according to S6K1 expression status.
In a multi-centre study conducted by van der Hage et al. in 2004 19 , the predictive potential of S6K1 upregulation was evaluated in premenopausal, node-negative early breast cancer patients. The authors showed that S6K1 upregulation was associated with poor loco-regional control in overall patients [hazard ratio (HR): 2.67, 95% confidence interval (CI): 1.39-5.14, p = 0.003] and in patients who received breast conserving surgery (HR 2.83, 95% CI: 1.42-5.62, p = 0.003), which comprised 80% of the total study population. Although it could be inferred that S6K1 upregulation is related to the response to radiotherapy because patients are likely to be treated with radiotherapy after breast conserving surgery, the role of radiotherapy was not examined in their study 19 . In another study conducted by Perez-Tenorio et al. in 2011 20 , S6K1 and S6K2 gene amplification was assessed in postmenopausal breast cancer patients. One of the aims of the study was to examine the clinical potential of both genes as predictive factors regarding radiotherapy, and the sub-analysis indicated that, in contrast to tumours with S6K1 amplification, tumours with a normal S6K1 gene copy number responded better to radiotherapy compared to chemotherapy (HR 0.27, 95% CI: 0.11-0.66, p = 0.0038) 20 . Although the authors suggested that S6 kinases may play a role in predicting the response to radiotherapy, the study could not compare radiotherapy versus none. In addition, a preclinical study conducted in 2006 supported a unique role of p-S6K1 in developing radio-resistance in breast cancer 18 . The authors showed that ionizing radiation increases both mTOR signalling and subsequent p-S6K1 levels in breast cancer cells. However, independent regulation of p-S6 ribosomal protein levels was observed over time despite mTOR activation 18 , implying that there may be other radiation-induced pathways www.nature.com/scientificreports www.nature.com/scientificreports/ regulating the phosphorylation of S6K1 12,27,28 . Consequently, in addition to simply acting as a downstream effector of the mTOR pathway, p-S6K1 may provide additional predictive information on radiation-mediated changes.
Our results provide additional evidence to support the aforementioned studies that the expression status of p-S6K1 could be a potential marker for radio-resistance in breast cancer. Although positive p-S6K1 expression appeared to be a better predictor of radio-resistance in patients with tumours >2 cm than in those with tumours ≤2 cm, we suggest that p-S6K1 status might nevertheless be associated with radio-resistance in tumours ≤2 cm based on the results of our experimental study. In addition, since our data were not derived from a matched set with respect to tumour size, the low number of recurrence events (n = 10) in patients with tumours ≤2 cm may have resulted in the insignificant p-value in this group. It is possible that the naturally excellent prognosis of this patient group (T1 treated with radiation therapy) was insufficient to detect a significant LRFS difference (5-year LRFS 98.7% vs. 97.4% for p-S6K1-negative and -positive patients, respectively; data not shown). Nevertheless, among the 10 recurrences, the proportion of tumours with positive p-S6K1 expression was exclusively higher (n = 9, 90%) than that for patients with a negative p-S6K1 status (10%), thereby supporting our hypothesis.
Moreover, although the role of p-S6K1 needs to be further explored, we suggest that the study of p-S6K1 may provide insights into the development of novel radio-sensitising agents in breast cancer patients. Increased recognition of the importance of the mTOR pathway in cancer treatment has called for exploration of mTOR inhibitors as radio-sensitisers for treating many types of cancer [29][30][31] . Nevertheless, studies on breast cancer in view of p-S6K1 are scarce. We found only one study in which pre-treatment with everolimus, a clinical mTOR inhibitor, radio-sensitised MDA-MB-231 and MCF7 breast cancer cells by blocking radiation-induced mTOR-S6K1 signalling 18 . The salient point of our experiments is that we chose to compare the expression levels of p-S6K1 in breast cancer stem cells with a radio-resistant phenotype (CD44 high /CD24 low MCF7 cells) to that of MCF7 breast cancer cells. Cancer stem cells are largely responsible for the development of radio-resistance during treatment because they have the potential to overcome the anti-cancer effects of radiotherapy and contribute to the generation of a radio-resistant heterogeneous tumour cell population [32][33][34] . The prominently increased p-S6K1 levels in radio-resistant breast cancer stem cells and distinct reversal of these levels by combining radiation and everolimus compared to control MCF cells indicate that p-S6K1 is at least partially responsible for treatment failure, and that inhibition of the mTOR-S6K1 pathway may provide an additional targeted approach to improve the outcome of radiotherapy.
A limitation of our work is that it was a single centre-based retrospective study. In addition, as there are no international standards guiding the staining and stratification of p-S6K1 expression levels, careful interpretation is required to avoid bias. The incidence of high p-S6K1 expression varies among studies, as different criteria have been used 16,19,[35][36][37] . Although upregulation of S6K1 expression is found in 10-30% of primary breast cancers, positive p-S6K1 expression was observed in approximately 75% of the samples in our study. Therefore, further prospective research in other populations is warranted to validate the general applicability of these results.
The strengths of our study are that we stratified relatively large numbers (n = 1770) of patients based on the receipt of radiotherapy to explicitly analyse the clinical association between p-S6K1 levels and radio-resistance. We demonstrated that p-S6K1could predict radio-resistance in breast cancer. In addition, this study is one of the few experimental studies to demonstrate the possibility of using S6K1 as a target for radio-sensitisation in breast cancer.
To date, several markers such as the cytoplasmic expression of peroxiredoxin-I and circulating tumour cells have been identified to correlate with the radiotherapy response in breast cancer 3,38,39 . However, most of these markers require a translational process for acquisition or interpretation and none is currently used to guide treatment decisions in clinical practice. Nevertheless, as p-S6K1 levels can be easily assessed with immunohistochemical staining owing to commercially available antibodies, they could be readily utilised in a clinical setting for predicting radio-resistance. A better understanding of the roles of p-S6K1 expression through experimental studies and accurate prediction of radio-resistance may provide a basis for its application in individualised radiotherapy in the future. conclusion p-S6K1 expression status was associated with a reduced response to radiotherapy in patients with breast cancer. This indicates the potential of p-S6K1 levels as a predictor of the response to radiotherapy and as a novel target for radio-sensitisation to increase the efficacy of radiotherapy.

Study design and patient selection.
A total of 1770 patients who received curative surgery for primary breast cancer between March 2008 and December 2015 were included in this study. Patients with in situ carcinoma and with a diagnosis of other primary malignancies were excluded. Radio-resistance was defined as having acquired loco-regional recurrence after completion of adjuvant radiotherapy. Information on patient age (<50 or ≥50 years), tumour size (<2 cm or ≥2 cm), nodal status (positive or negative), hormone receptor status, human epidermal growth factor receptor (HER)-2 status, histologic grade (1 and 2 or 3), type of surgery, and p-S6K1 status were reviewed retrospectively from our database, a web-based system that has been used to collect information on breast cancer patients since 1983. To date, more than 17,000 breast cancer patients have been registered in this database. The requirement for informed consent was waived owing to the retrospective nature of this study by the institutional review board of Korea Cancer Centre Hospital that approved the protocols of this study www.nature.com/scientificreports www.nature.com/scientificreports/ Immunohistochemical staining. Routine immunohistochemical assessment of oestrogen receptor (ER), progesterone receptor (PR), HER-2, and p-S6K1 expression was performed after the acquisition of each specimen at the diagnosis of breast cancer prior to surgery. Formalin fixed, paraffin-embedded tumour tissue blocks of core needle biopsied specimens were preferentially used 41 . However, tissue from post-surgical specimens was used when core needle biopsied specimens were unavailable. ER positivity was determined as the expression of ER detected in at least 1% of tumour cells as determined by immunohistochemistry. HER-2 upregulation was determined as a score of 3+ in immunohistochemical staining or a score of 2+ with positive gene amplification in in situ hybridisation. The p-S6K1 expression status was evaluated with immunohistochemistry and scored in a range from 0 to 3+, with a mouse monoclonal antibody against p-S6K1 (Cell Signaling Technology, Danvers, MA, USA; dilution 1:50). A score of 0 was defined as p-S6K1-negative, whereas scores from 1+ to 3+ were considered as p-S6K1-positive, with higher values indicating increased expression levels of p-S6K1. Details of the procedures are described in our previous report, in which we successfully evaluated the p-S6K1 status of 304 breast cancer patients 17  Chi-square and Fisher's exact tests were employed to investigate the correlation between the clinico-pathological parameters and p-S6K1 expression status in each group. LRFS was defined as the time from the diagnosis of primary breast cancer to the time of first detection of loco-regional recurrence by physical examination or radiological imaging. Loco-regional recurrences presenting simultaneously with or after the diagnosis of distant metastasis were included. Kaplan-Meier plots with log-rank tests were used to analyse differences in survival between groups. Univariate and multivariate analyses were performed with the Cox-proportional hazard regression model to calculate HR and 95% CI values. A p-value of <0.05 was considered statistically significant.

Data availability
Most of the data generated or analysed during this study are included in this published article and supplementary materials. The datasets generated during the study are available from the corresponding author on reasonable request.