High‐level expression of ARID1A predicts a favourable outcome in triple‐negative breast cancer patients receiving paclitaxel‐based chemotherapy

Abstract Paclitaxel‐based chemotherapy is a common strategy to treat patients with triple‐negative breast cancer (TNBC). As paclitaxel resistance is still a clinical issue in treating TNBCs, identifying molecular markers for predicting pathologic responses to paclitaxel treatment is thus urgently needed. Here, we report that an AT‐rich interaction domain 1A (ARID1A) transcript is up‐regulated in paclitaxel‐sensitive TNBC cells but down‐regulated in paclitaxel‐resistant cells upon paclitaxel treatment. Moreover, ARID1A expression was negatively correlated with the IC 50 concentration of paclitaxel in the tested TNBC cell lines. Kaplan‐Meier analyses revealed that ARID1A down‐regulation was related to a poorer response to paclitaxel‐based chemotherapy in patients with TNBCs as measured by the recurrence‐free survival probability. The pharmaceutical inhibition with p38MAPK‐specific inhibitor SCIO‐469 revealed that p38MAPK‐related signalling axis regulates ARID1A expression and thereby modulates paclitaxel sensitivity in TNBC cells. These findings suggest that ARID1A could be used as a prognostic factor to estimate the pathological complete response for TNBC patients who decide to receive paclitaxel‐based chemotherapy.


| INTRODUCTION
Breast cancer is a major public health problem in females worldwide. 1 The universality and occurrence rate of breast cancer have risen markedly over the past several decades. Of note, triple-negative breast cancer (TNBC), which lacks the characteristic receptors for oestrogen, progesterone and Her2/neu, is an aggressive tumour that is associated poor survival and represents an important clinical challenge. 2 Currently, chemotherapy remains the standard breast cancer treatment. However, breast cancers may display chemoresistance and radioresistance. 3 Paclitaxel, which is an extract of the tree Taxus brevifolia, is a potent chemotherapeutic agent used against breast cancers. 4 Paclitaxel mechanistically stabilizes tubulin polymerization resulting in arrest of mitosis and subsequent apoptosis. 5 However, paclitaxel has had limited success in cancer therapy because of the activation of cytoprotective signalling pathways, including the nuclear factor kappa B (NF-jB), phosphatidylinositol 3-kinase (PI3K)/Akt and mitogen-activated protein kinase (MAPK) signalling pathways, which induce drug resistance. [6][7][8]  proteins. 9 Chromatin remodelling, which regulates the synthesis, transcription and repair of DNA, is important in cell nuclear activities. Genetic mutation of the chromatin remodelling complex has been identified as a mechanism of tumour occurrence and development. 10 Here, we analyse the transcriptional profiling of paclitaxel-sensitive DU4475 and paclitaxel-resistant MDA-MB436 without or with paclitaxel treatment, which was determined previously. 11 We found that the AT-rich interaction domain 1A (ARID1A) is up-regulated in DU4475 cells but is down-regulated in MDA-MB436 cells. ARID1A is a noncatalytic subunit of the chromatin remodelling complex and has the ability to combine with DNA or proteins. 12 Previous studies have demonstrated that ARID1A is a tumour suppressor that is frequently mutated in various cancers, including breast cancer. [13][14][15] ARID1A loss correlates with mismatch repair deficiency and intact p53 expression in endometrial cancer. 16 Recently, genetic mutations of ARID1A have been shown to be associated with treatment and prognosis of the tumour. 12 Mamo et al indicated that low ARID1A RNA or protein expression is related with more aggressive breast cancers. 15 However, whether such a high mutation rate is associated with the resistance of breast cancer to chemotherapy remains unclear and requires further investigation. Therefore, the aims of this study were to evaluate the effect of the ARID1A gene on breast cancer following paclitaxel treatment and to investigate the possible mechanism.
The results suggest that ARID1A may be used to predict the outcome in breast cancer patients receiving paclitaxel-based chemotherapy. ARID1A thus warrants further investigation as a potential diagnostic and therapeutic marker for breast cancer.

| Western blot analysis
The protein concentrations of total cell lysates, nuclear and cytoplasmic extracts were determined by the Bradford assay (Bio-Rad, Hercules, CA, USA) using bovine serum albumin as a standard. Samples containing equal quantity of proteins were mixed in Laemmli sample buffer (62.5 mmol/L Tris [pH 6.7], 1.25% SDS, 12.5% glycerol, and 2.5% b-mercaptoethanol) and boiled for 10 minutes at 100°C before being separated by electrophoresis on 8%-15% SDS-PAGE gels and transferred to polyvinylidene difluoride membrane (Millipore, Temecula, CA, USA). After blocking with 5% dry milk in PBST, membranes were explored with antibodies against p-p38/p38/ARID1A (Cell Signaling, Danvers, MA, USA) and GAPDH (AbFrontier, Seoul, Korea).
The horseradish peroxidase-conjugated secondary antibody incubation was performed, and the specific immunoreactive protein complexes were detected by using the enhanced chemiluminescense method (Amersham Bioscience, Tokyo, Japan).

| Univariate and multivariate analyses
The 400 breast cancer patients who received post-operative chemotherapy from the TCGA database were used to perform univariate and multivariate analyses using Cox regression tests. ARID1A expression levels and clinical data, including age, pathological stage, T and N, were input as variables for the Cox regression test using RFS conditions. Mann-Whitney U-tests were used to analyse non-parametric data.  Figure 1A). We found 93 consensus genes with 1.5-fold changes after paclitaxel treatment in DU4475 and MDA-MB436 cells (Table S1). Our data showed that  (Figures 2A and S2A-C).

| Establishment of meta-analysis
Whereas the mRNA levels of VMP1/MIR21 detected by probe 224917_at were shown to be positively correlated, the VMP1/ MIR21 mRNA levels detected by probe 220990_s_at were negatively correlated with the paclitaxel IC 50 concentrations of each tested cell line (Figures 2B and S2D). However, the changes of VMP1/MIR21 mRNA after paclitaxel treatment were inversely correlated with the respective paclitaxel IC 50 concentrations in the detected breast cancer cells (Figures 2B and S2D). Because

| ARID1A expression is predominantly downregulated in the majority of basal-like breast cancer tissues
We next dissected the transcriptional profiling of ARID1A across all breast cancer subtypes using the TCGA database. 18 We made use of Agilent microarray and RNA sequencing to analyse ARID1A gene

| ARID1A down-regulation predicts a significantly shorter RFS in breast cancer
To further validate the potential prognostic significance of our findings, we researched publicly available platforms of expression analysis, including SurvExpress, 19 K-M Plotter 20 and PrognoScan. 17 First, we estimated the prognostic significance of ARID1A in predicting the RFS rates, which frequently reflect chemotherapeutic responses in breast cancer patients. From the SurvExpress database, the prognostic value of low ARID1A mRNA expression was significantly correlated with poor RFS rates in 1901 breast cancer patients ( Figure 4A). Furthermore, the mRNA levels of ARID1A in the high-risk cohort were significantly down-regulated compared to the low-risk cohort in breast cancer patients ( Figure 4A). Accordingly, ARID1A down-regulation appeared to be associated with unfavourable RFS rates in breast cancer patients, based on the K-M Plotter database ( Figure 4B). Similar results were also observed in different ARID1A probes within the K-M Plotter database against breast cancer patients ( Figure S3). We found that patients with higher ARID1A expression levels were more likely to have a favourable RFS rate. Next, we performed a global meta-analysis for ARID1A using the PrognoScan database. Using a Cox P-value of <.05 ( Figure 4C), lower ARID1A expression in blood, breast, lung and colorectal cancers, with the exception of soft tissue cancer, was associated with a poorer outcome. Particularly in breast cancer, ARID1A down-regulation was highly correlated with cancer progression, for example, metastasis and recurrence ( Figure 4C).
Using TCGA database, we found that ARID1A down-regulation refers to a poor RFS probability in unclassified breast invasive carcinoma (BRCA) patients ( Figure 4D). Significantly, ARID1A down-regulation was appeared to strongly predict an unfavourable RFS rate in patients with TNBC in TCGA database ( Figure 4E).  Table 1). In addition, we analysed ARID1A expression in breast cancer patients receiving paclitaxel-based chemotherapy. The results revealed that the expression in unclassified BRCA and TNBC patients with cancer recurrence was significantly lower than that in patients with no recurrence ( Figure 5D). Neoadjuvant chemotherapy (NAC) is used in breast cancer treatment of downstage tumours. 21 We analysed the transcriptional profile of ARID1A in breast tumours derived from breast cancer patients receiving preoperative paclitaxel-based NAC by using GSE22513 data set. [22][23][24] ARID1A mRNA levels in breast tumours derived from patients with no pathological F I G U R E 5 ARID1A down-regulation predicts a poor response to paclitaxel (PTX) chemotherapy in breast cancer patients. (A-C) Kaplan-Meier analysis for ARID1A expression under the condition of RFS probability in unclassified breast cancer patients and TNBC cohort with postoperative chemotherapy using SurvExpress (A) and TCGA (B and C) databases. (D) Boxplot for mRNA levels of ARID1A in tumour biopsy derived from unclassified breast cancer patients and TNBC cohort receiving post-operative chemotherapy without (No) or with (Yes) cancer recurrence using TCGA database. The statistical difference was analysed by t test. (E) Boxplot for mRNA levels of ARID1A in tumour biopsy derived from breast cancer patients with PTX pretreatment using GSE22513 data set. The statistical difference was estimated by nonparametric Mann-Whitney test. (F) Boxplot for mRNA levels of ARID1A in breast cancer tissues derived from ER(+), HER2(+) and triplenegative, which identified by immunohistochemistry (IHC) analysis, breast cancer patients who received pre-operative chemotherapy (prechemo.) using GSE32646 data set. The statistical differences were analysed by one-way ANOVA using Turkey's test. In (E) and (F) pCR and nCR denote pathological complete response and no pathological complete response, respectively complete response (nCR) were significantly lower than those from patients with pathological complete response (pCR) ( Figure 5E).
Moreover, using GSE32646 data set, 25 among patients receiving paclitaxel-based NAC, ARID1A mRNA levels in TNBC, but not breast tumour of ER(+) and HER2(+), derived from patients with nCR were significantly (P < .05) decreased compared to those of patients with pCR ( Figure 5F).

| p38 mitogen-activated protein kinase-related pathways underlying paclitaxel resistance in TNBC cells
We next performed an in silico analysis using IPA software to predict potentially activated/inhibited upstream regulators related to the mechanism of paclitaxel resistance in breast cancer cells. As shown in Figure 6A and Table S2, p38 mitogen-activated protein kinase (p38MAPK) and hydrogen peroxide were up-regulated in DU4475 cells but down-regulated in MDA-MB436 cells. Using TCGA database, we found that ARID1A mRNA levels in TNBCs positively correlate with the protein levels of phosphorylated p38MAPK (pp38MAPK) which is known as an activated protein form of p38MAPK ( Figure 6B). Moreover, our data showed that the protein levels of pp38MAPK in a panel of TNBC cell lines ( Figure 6C) inversely correlate with paclitaxel IC50 concentrations ( Figure 6D). Significantly, the pretreatment with p38MAPK inhibitor SCIO-469 26 dose dependently suppressed the enhanced ARID1A protein levels by paclitaxel in HCC70 cells that are relative sensitive to paclitaxel treatment ( Figure 6E). In contrast, the pretreatment of SCIO-469 significantly (P < .05) promoted the paclitaxel resistance in HCC70 cells ( Figure 6F).

| DISCUSSION
Chemoresistance, the main obstacle in cancer therapy, is caused by the onset of drug-resistant cells in cancer tissues. 27 Currently, the standard treatment for TNBCs is cytotoxic chemotherapy. However, some patients with TNBCs are highly chemotherapy resistant and relapse quickly after treatment in the adjuvant setting. 21 Paclitaxel is the first-line chemotherapeutic drug for clinical treatment of TNBCs.
However, drug resistance often appears chemotherapy. 28  shown to be a useful biomarker to predict neoadjuvant therapeutic response in breast cancer patients. 32 Evidence indicates that partial loss of ARID1A expression is significantly correlated with poor disease-free survival in patients with invasive breast carcinoma. ARID1A down-regulation significantly upregulated RAB11 family interacting protein 1 (RAB11FIP1) mRNA in breast cancer cells. 33 RAB11FIP1 is known as a Rab-coupling protein that assists breast cancer progression. 34 In the present study, ARID1A expression was up-regulated in TNBCs compared to other subtypes derived from patients with breast cancer (Figure 3A-D). ARID1A down-regulation refers to a poor RFS rate in unclassified BRCA and TNBC patients (Figures 4 and S3). Furthermore, the mRNA levels of ARID1A in the high-risk cohort are significantly down-regulated compared to the low-risk cohort in unclassified BRCA patients ( Figure 4A). We also performed in silico analysis using IPA software to predict potentially activated/inhibited upstream regulators in TNBC cells. We found that p38MAPK and hydrogen peroxide were up-regulated in DU4475 cells but were down-regulated in MDA-MB436 cells ( Figure 6A and Table S2). The p38MAPK signalling pathway induces cell activation, proliferation and apoptosis. 35 Recent studies have shown that the p38MAPK pathway is closely associated with drug resistance in cancer therapy. Sanchez-Prieto et al indicated that inhibition of p38MAPK decreased the apoptotic fraction of cells exposed to chemotherapeutic agents and increased cell survival. 36 Another recent study concluded that p38MAPK inhibition blocked p53-dependent apoptosis allowing an autophagic response that mediated resistance. 37 Previous studies have demonstrated that the antioxidant capacity of tumour cells scavenges excessive reactive oxygen species (ROS), allowing the disease to progress and develop resistance to apoptosis. 38 Furthermore, toxic levels of ROS produced in cancers are anti-tumorigenic, resulting in an increase in oxidative stress and induction of tumour cell death. 39,40 Here, we showed that ARID1A expression was positively correlated with the protein levels of pp38MAPK and highly regulated by p38MAPK-related pathways in TNBCs.
In conclusion, the results of the present study demonstrated that

CONF LICT OF I NTEREST
The authors declare no conflict of interest.