Chemotherapeutic response is critical for the successful treatment and good prognosis in cancer patients. In this study, we analysed the gene expression profiles of preoperative samples from oestrogen receptor (ER)-negative breast cancer patients with different responses to taxane-anthracycline-based (TA-based) chemotherapy, and identified a group of genes that was predictive. Pregnancy specific beta-1-glycoprotein 1 (PSG1) played a central role within signalling pathways of these genes. Inhibiting PSG1 can effectively reduce chemoresistance via a transforming growth factor-β (TGF-β)-related pathway in ER-negative breast cancer cells. Drug screening then identified dicumarol (DCM) to target the PSG1 and inhibit chemoresistance to TA-based chemotherapy in vitro, in vivo, and in clinical samples. Taken together, this study highlights PSG1 as an important mediator of chemoresistance, whose effect could be diminished by DCM.

CLINICAL PERSPECTIVES

  • Chemoresistance is one of the central hindrances to cancer therapy.

  • It is important to find more agents that can improve the efficacy of chemotherapy. In this study, we identified the FDA-approved agent dicumarol could inhibit the chemoresistant feature of breast cancer cells.

  • Thus our results may help to discovery effect anti-chemoresistance agent and methods clinically.

INTRODUCTION

Oestrogen receptor (ER)-negative breast cancers contain a distinct group of cancer cells compared with ER-positive breast cancers. They are more aggressive than ER-positive breast cancers, most of them are insensitive to anti-oestrogen therapy, and cytotoxic chemotherapy is currently accepted as the standard of care [1]. Of the multiple combinations of chemotherapeutic drugs, taxane-anthracycline-based (TA-based) chemotherapy remains the cornerstone of current treatments for ER-negative breast cancers [25]. However, due to the aggressive and heterogeneous features of these cancers, the rate of pathologically complete response for TA-based chemotherapy is still <30%, and a substantial proportion of patients are chemoresistant and suffer from residual disease after chemotherapy [3,5], which can compromise a better prognosis [6]. Therefore, a means of predicting the sensitivity to TA-based chemotherapy would help patients to choose personalized cancer therapy and improve both the efficiency and effect of the selected therapy.

Gene expression microarrays are now widely used clinically and various gene expression features have been found to assess the biological effect and risk stratification of breast cancers [79]. Furthermore, these features are commonly represented by a list of significantly ectopic genes in cancers, which can provide therapeutic targets in cancer treatment. However, few studies have specifically deciphered the gene expression features in ER-negative breast cancers with TA-based chemotherapy.

Therefore, this study aimed to identify the gene expression features of chemosensitive and chemoresistant ER-negative breast cancer patients under the TA-based chemotherapeutic strategy, and found that pregnancy specific beta-1-glycoprotein 1 (PSG1) played a central role. We further demonstrated that dicumarol (DCM) targeted PSG1 and enhanced the cytotoxicity of taxanes and anthracyclines. Therefore, the results of our study can help ER-negative breast cancer patients to achieve greater clinical benefits from TA-based chemotherapy.

MATERIALS AND METHODS

Detailed methods are described in the supplementary material.

Preparation of expression data

Publicly available datasets of breast cancers in the Gene Expression Omnibus were used (Supplementary Table S1). Microarray Suite version 5.0 (MAS 5.0) was used to normalize the microarray data in the regressive method, as described in a related study [8]. After separating the patients into chemoresistant or sensitive groups according to the standards indicated below, differences in gene expression were calculated by one-way ANOVA.

Collagen-gel droplet embedded-culture drug-sensitivity test

Each patient involved in this study provided written informed consent. All protocols for enrolling patients and sample analysis were conducted with the approval of the Ethics Committees of the university and hospital. Collagen-gel droplet embedded-culture drug-sensitivity tests (CD-DSTs) were performed as in previous studies by others and ourselves [10,11]. Detailed descriptions of these protocols are available in the supplementary material.

RESULTS

Nineteen genes are ectopically expressed and are predictive in ER-negative patients

The GSE25055, GSE25065 and GSE41998 datasets together were chosen to form the discovery cohorts (Supplementary Table S1). The ER-negative patients were grouped based on their response to TA-based chemotherapy as suggested in previous studies [8,12]. Patients with a pathologically complete response were defined as chemosensitive (n=70), whereas those with an extensive residual cancer burden (RCB-III; GSE25055 and 25065) or residual disease (GSE41998) were considered to be chemoresistant (n=84). Nineteen genes differed significantly between chemoresistant and sensitive ER-negative breast cancer patients (Figure 1A and Supplementary Table S2) and thus were chosen to generate a signature for discriminating responses to TA-based chemotherapy (termed TASig). According to the previous studies, 14 of the 19 genes were associated with hallmarks of cancer [13] (Supplementary Table S2), so significant changes in their expression may contribute to the progression of cancer and the development of chemoresistance.

PSG1 plays a central role in TASig

Figure 1
PSG1 plays a central role in TASig

(A) The heat map depicts the two-way hierarchical clustering of ER-negative breast tumour samples with 19 genes in TASig. Seventy chemoresistant and 84 chemosensitive patients from GSE25055, 25065 and 41998 were included. (B) Genetic interaction network of TASig. Genes in the TASig, as well as genes that significantly change between chemoresistant and chemosensitive patients in at least two of the datasets from GSE25055, 25065 and 41998 were analysed. (C) TASig status in MCF-7/MDR cells. The gene expression of TASig in MCF-7/MDR, chemosensitive control MCF-7/WT and PSG1 siRNA (siPSG1) treated MCF-7/MDR cells were analysed by real-time-PCR. Mean values were shown in heat map. Arrows indicate significantly changed genes in MCF-7/WT and siPSG1 treated MCF-7/MDR cells compared with MCF-7/MDR cells. (D) PSG1 expression in MCF-7/MDR and MCF-7/WT cells was analysed by western blot. (E) Left panel: knocking down of PSG1 reversed the chemoresistance. MCF-7/MDR cells were transfected with either siPSG1 or a scrambled siRNA, then treated with either PTX or ADM (20 μg/ml). The cells that were transfected with scrambled siRNA and treated with PTX or ADM were used as control, whose cell viability was normalized to 100%. *P<0.05 compared with control. Right panel: recombinant PSG1 rescued the effect of PSG1 siRNA. MCF-7/MDR cells were transfected with PSG1 siRNA, treated with PTX or ADM with (si+PSG1) or without (siPSG1) recombinant PSG1. MCF-7/MDR cells that were transfected with scrambled siRNA, treated with PTX or ADM were used as control, whose cell viability was normalized to 100%.

Figure 1
PSG1 plays a central role in TASig

(A) The heat map depicts the two-way hierarchical clustering of ER-negative breast tumour samples with 19 genes in TASig. Seventy chemoresistant and 84 chemosensitive patients from GSE25055, 25065 and 41998 were included. (B) Genetic interaction network of TASig. Genes in the TASig, as well as genes that significantly change between chemoresistant and chemosensitive patients in at least two of the datasets from GSE25055, 25065 and 41998 were analysed. (C) TASig status in MCF-7/MDR cells. The gene expression of TASig in MCF-7/MDR, chemosensitive control MCF-7/WT and PSG1 siRNA (siPSG1) treated MCF-7/MDR cells were analysed by real-time-PCR. Mean values were shown in heat map. Arrows indicate significantly changed genes in MCF-7/WT and siPSG1 treated MCF-7/MDR cells compared with MCF-7/MDR cells. (D) PSG1 expression in MCF-7/MDR and MCF-7/WT cells was analysed by western blot. (E) Left panel: knocking down of PSG1 reversed the chemoresistance. MCF-7/MDR cells were transfected with either siPSG1 or a scrambled siRNA, then treated with either PTX or ADM (20 μg/ml). The cells that were transfected with scrambled siRNA and treated with PTX or ADM were used as control, whose cell viability was normalized to 100%. *P<0.05 compared with control. Right panel: recombinant PSG1 rescued the effect of PSG1 siRNA. MCF-7/MDR cells were transfected with PSG1 siRNA, treated with PTX or ADM with (si+PSG1) or without (siPSG1) recombinant PSG1. MCF-7/MDR cells that were transfected with scrambled siRNA, treated with PTX or ADM were used as control, whose cell viability was normalized to 100%.

The 19-gene TASig was first trained on GSE25055 using a Bayesian discriminative method that could quantitatively assess the chemoresistant phenotype as we demonstrated previously [14,15]. The TASig score for each patient in GSE25055 was calculated from a discriminative equation (Supplementary Tables S3A and S3B). Thereafter, the negative predictive value (NPV) and positive predictive value (PPV) of the equation were 84.4% and 93.8% respectively (Table 1). The accuracy of the equation in predicting NPV and PPV was high as calculated by the AUC (area under the curve), which was 0.981 (Supplementary Figure S1A).

Table 1
Performance of TASig and its equation for predicting resistance and sensitivity to TA-based chemotherapy in ER-negative breast cancer patients

*NPV, predicted to be sensitive/total % sensitive; PPV, predicted to be resistant/total % resistant; confidence intervals calculated by binomial approximation.

Training datasetValidation datasets
GSE25055GSE25065GSE41998GSE20194
NPV* PPV NPV PPV NPV PPV NPV PPV 
84.4 93.8 90.9 100 85.7 82.1 85.7 79.1 
(70–93) (79–99) (58–99) (74–100) (67–96) (63–94) (67–96) (64–90) 
Training datasetValidation datasets
GSE25055GSE25065GSE41998GSE20194
NPV* PPV NPV PPV NPV PPV NPV PPV 
84.4 93.8 90.9 100 85.7 82.1 85.7 79.1 
(70–93) (79–99) (58–99) (74–100) (67–96) (63–94) (67–96) (64–90) 

Furthermore, the discriminative power of TASig was validated in GSE25065 and GSE41998 from the discovery cohorts, as well as in another independent cohort, GSE20194. The TASig scores of patients from these datasets were calculated based on their microarray gene expression data, and they were grouped later according to the scores and the cutoff value. As a result, the equation showed good prediction of both chemoresistant and sensitive patients (Table 1).

Central role of PSG1

Based on the fact that TASig could effectively discriminate chemoresistant patients from chemosensitive ones, we further tried to identify the molecular mechanism of TASig in mediating chemoresistance. We constructed a genetic interaction network of TASig using Cytoscape 3.2.0 [16] plugin GeneMANIA [17]. PSG1 appeared in the central location of the network (Figure 1B). The predictive and prognostic power of PSG1 was then analysed. The median PSG1 value of patients in each dataset was used to separate chemoresistant and sensitive groups [18,19]; patients with higher-PSG1 value were grouped as chemoresistant, and lower ones were chemosensitive. The predictive PPV and NPV of PSG1 were calculated in GSE25055, 25065, 41998 and 20194, and data are shown in Table 2. The accuracy of PSG1 in predicting NPV and PPV was calculated by the AUC, which was 0.969 (Supplementary Figure S1B). In GSE25055 and 25065 which recorded distance relapse free survival (DRFS) status of patients, the DRFS events of PSG1-defined chemoresistant and sensitive patients were calculated by the Kaplan–Meier method. As shown in Supplementary Figure S1(C), patients predicted to be chemoresistant showed significantly worse outcomes than sensitive ones. The prognostic power of the TASig was then analysed in datasets without chemotherapy on ER-negative breast cancer patients. In these datasets, the patients were grouped according to the median PSG1 expression value, and DRFS events were calculated between higher-PSG1 patients with lower-PSG1 individuals. As shown in Supplementary Figure S1(D), PSG1 only showed good prognostic power in dataset GSE22220 [20], but not in GSE33926 and GSE22226 [21,22].

Table 2
Predictive effect of PSG1
Training datasetValidation datasets
GSE25055GSE25065GSE41998GSE20194
NPV PPV NPV PPV NPV PPV NPV PPV 
71.1 71.9 72.7 70.0 82.1 78.6 75.0 79.1 
(58–84) (56–87) (46–99) (41–98) (68–96) (63–94) (59–91) (67–91) 
Training datasetValidation datasets
GSE25055GSE25065GSE41998GSE20194
NPV PPV NPV PPV NPV PPV NPV PPV 
71.1 71.9 72.7 70.0 82.1 78.6 75.0 79.1 
(58–84) (56–87) (46–99) (41–98) (68–96) (63–94) (59–91) (67–91) 

We next analysed the molecular mechanism of PSG1 in chemoresistance with the breast cancer cell line MCF-7/MDR, which are resistant to both adriamycin (ADM) and paclitaxel (PTX) [23] and suitable for drug screening for two reasons: first, MCF-7/MDR is ER-negative (Supplementary Figure S2A); second, real-time PCR showed that 11 genes, including PSG1 in MCF-7/MDR cells had a trend of change similar to that of TASig genes in chemoresistant patients compared with chemosensitive parental control MCF-7/WT cells (Figure 1C and Supplementary Figure S2B, western blot for PSG1 in Figure 1D).

The effect of PSG1 in chemoresistance was then studied. When the PSG1 gene was knocked down by siRNA (siPSG1; Supplementary Figure S2C) in MCF-7/MDR cells, expression of six genes in TASig was significantly reversed by siPSG1 (Figure 1C arrows, and Supplementary Figure S2B), again suggesting PSG1 may control the functions of other genes in TASig, which needs further validation. Then siPSG1-transfected MCF-7/MDR cells were treated with ADM or PTX. The results showed that, without PSG1 production, more MCF-7/MDR cells were killed by ADM or PTX (Figure 1E). Then the cells transfected with siPSG1 were rescued by adding recombinant PSG1 protein (Abnova), and their chemoresistance was regained (Figure 1E). These results suggested that PSG1 mediates chemoresistance in MCF-7/MDR cells.

Dicumarol modulates PSG1

The fact that the TASig was based on the common denominator of gene expression in TA-resistant patients provided a great opportunity to overcome the resistance by finding active drugs that target this signature. We then screened the biologically active small molecule library in our university for both chemoresistance reversing and PSG1 modulating molecules. The screened molecules have possible anti-cancer activity either reported by previous studies or by preliminary tests of the university (Supplementary Table S4). We later identified DCM as the most potent one. DCM reversed the PSG1 expression and extracellular secretion (Figure 2A). DCM was then found to inhibit chemoresistance of MCF-7/MDR cells. At 20 μg/ml, the inhibitory ratio (IR) of ADM was 20±3.2% and that of PTX was 12±1.4% in MCF-7/MDR cells. However, when combined with DCM at 12.5 μg/ml, the IR increased significantly to 83±3.8% for ADM and 64±4.6% for PTX (Figure 2B). In addition, when the MCF-7/MDR cells were treated with ADM in combination with PTX, DCM could increase the IR of the drugs from 31±3.5% to 96±0.6% (Figure 2B). Furthermore, when recombinant PSG1 compensated for the PSG1 loss caused by DCM treatment, the cytotoxicity of ADM plus DCM, or PTX plus DCM in MCF-7/MDR cells was diminished (Figure 2C).

DCM inhibits chemoresistance via PSG1

Figure 2
DCM inhibits chemoresistance via PSG1

(A) Cells were treated with 12.5 μg/ml DCM, and expression and extracellular level of PSG1 was analysed by Western blot (left) or ELISA (right). *P<0.05, **P<0.01 compared with MCF-7/MDR that was normalized to 1. (B) DCM increases the cytotoxicity of ADM and PTX. Cytotoxicity of ADM (left panel) or PTX (middle panel) alone (20 μg/ml) was analysed by MTT and shown as inhibition ratio in bar graph. The cells were also treated with different concentration of DCM together with ADM (left panel) or PTX (middle panel) (20 μg/ml), the inhibition ratio was non-liner regressed. The red dots represent cell viability when DCM was 12.5 μg/ml. Right panel: Cell viability was also analysed in MCF-7/MDR cells treated with ADM and PTX together (A.P) or ADM, PTX and DCM together (A.P.D). (C) Recombinant PSG1 reversed toxicity of DCM. MCF-7/MDR cells were treated with ADM+DCM or PTX+DCM in the presence or absence (control, Ctrl) of PSG1. Cell viability of control was normalized to 100%. *P<0.05 compared with control.

Figure 2
DCM inhibits chemoresistance via PSG1

(A) Cells were treated with 12.5 μg/ml DCM, and expression and extracellular level of PSG1 was analysed by Western blot (left) or ELISA (right). *P<0.05, **P<0.01 compared with MCF-7/MDR that was normalized to 1. (B) DCM increases the cytotoxicity of ADM and PTX. Cytotoxicity of ADM (left panel) or PTX (middle panel) alone (20 μg/ml) was analysed by MTT and shown as inhibition ratio in bar graph. The cells were also treated with different concentration of DCM together with ADM (left panel) or PTX (middle panel) (20 μg/ml), the inhibition ratio was non-liner regressed. The red dots represent cell viability when DCM was 12.5 μg/ml. Right panel: Cell viability was also analysed in MCF-7/MDR cells treated with ADM and PTX together (A.P) or ADM, PTX and DCM together (A.P.D). (C) Recombinant PSG1 reversed toxicity of DCM. MCF-7/MDR cells were treated with ADM+DCM or PTX+DCM in the presence or absence (control, Ctrl) of PSG1. Cell viability of control was normalized to 100%. *P<0.05 compared with control.

Furthermore, DCM showed small cytotoxicity to non-cancerous cells. The IRs of DCM (12.5 μg/ml) were 4±0.8% and 8±1.2% in normal breast epithelial cell line MCF-7/10A and primary breast epithelial cells (HMEC) respectively (Supplementary Figure S3A). The IR of DCM (12.5 μg/ml) alone in MCF-7/MDR cells was 23±2.3% but still lower when combined with ADM and PTX (Supplementary Figure S3B).

Mechanism of DCM in modulating PSG1

The upstream component of DCM regulation of PSG1 was first analysed. Chromatin immunoprecipitation assay (ChIP) sequencing datasets in UCSC website indicated that the MYC may bind to PSG1 promoter region, chr19:43353148–43353203. We therefore validated such binding in MCF-7/MDR cells by ChIP (Figure 3A). Knocking down of MYC decreased PSG1 expression in MCF-7/MDR cells (Figure 3B). When treated with DCM, the binding activity of MYC significantly decreased (Figure 3A). Therefore, the results suggest DCM works on MYC to modulate PSG1.

Mechanism of DCM to reverse chemoresistance

Figure 3
Mechanism of DCM to reverse chemoresistance

(A) ChIP assay analysis the MYC binding on PSG1 promoter. The normal rabbit IgG was used as control. (B) MYC siRNA (siRNA) inhibited the PSG1 expression. (C) PSG1 activates TGF-β1. Active form of TGF-β1 was analysed by ELISA in supernatant of MCF-7/WT cells, MCF-7/MDR cells, PSG1 siRNA transfected or DCM treated (12.5 μg/ml) MCF-7/MDR cells. Scrambled siRNA was used as control, whose level of active TGF-β1 was normalized to 1. *P<0.05, **P<0.01 compared with control. (D) PSG1 siRNA transfected or DCM treated MCF-7/MDR cells were treated with different concentration of recombinant PSG1 protein, and the active form of TGF-β1 was analysed by ELISA in supernatant of MCF-7/MDR cells. **P<0.01 compared with control. (E) TGF-β1 activation was analysed by luciferase activity of SBE reporter in MCF-7/MDR cells. PSG1 untreated cells were used as control and its luciferase activity was normalized to 1. **P<0.01 compared with control. (F) Knocking down of TGF-β1 reversed the chemoresistance. MCF-7/MDR cells were transfected with either TGF-β1 siRNA or a scrambled siRNA, then treated with either PTX or ADM (20 μg/ml). The cells that were transfected with scrambled siRNA and treated with PTX or ADM were used as control, whose cell viability was normalized to 100%. **P<0.01 compared with control. (G) DCM inhibits EMT pathway. MCF-7/MDR cells were treated with DCM in the presence or absence of recombinant PSG1, then the expression of EMT hall markers were analysed by western blot.

Figure 3
Mechanism of DCM to reverse chemoresistance

(A) ChIP assay analysis the MYC binding on PSG1 promoter. The normal rabbit IgG was used as control. (B) MYC siRNA (siRNA) inhibited the PSG1 expression. (C) PSG1 activates TGF-β1. Active form of TGF-β1 was analysed by ELISA in supernatant of MCF-7/WT cells, MCF-7/MDR cells, PSG1 siRNA transfected or DCM treated (12.5 μg/ml) MCF-7/MDR cells. Scrambled siRNA was used as control, whose level of active TGF-β1 was normalized to 1. *P<0.05, **P<0.01 compared with control. (D) PSG1 siRNA transfected or DCM treated MCF-7/MDR cells were treated with different concentration of recombinant PSG1 protein, and the active form of TGF-β1 was analysed by ELISA in supernatant of MCF-7/MDR cells. **P<0.01 compared with control. (E) TGF-β1 activation was analysed by luciferase activity of SBE reporter in MCF-7/MDR cells. PSG1 untreated cells were used as control and its luciferase activity was normalized to 1. **P<0.01 compared with control. (F) Knocking down of TGF-β1 reversed the chemoresistance. MCF-7/MDR cells were transfected with either TGF-β1 siRNA or a scrambled siRNA, then treated with either PTX or ADM (20 μg/ml). The cells that were transfected with scrambled siRNA and treated with PTX or ADM were used as control, whose cell viability was normalized to 100%. **P<0.01 compared with control. (G) DCM inhibits EMT pathway. MCF-7/MDR cells were treated with DCM in the presence or absence of recombinant PSG1, then the expression of EMT hall markers were analysed by western blot.

The downstream component of DCM regulation on PSG1 was then analysed. PSG1 has been shown to activate extracellular transforming growth factor-β (TGF-β) [24,25], and autocrine or paracrine TGF-β1 is an important contributor to cancer progression, including chemoresistance [26]. Extracellular active TGF-β1 was analysed by ELISA that only recognizes the active form, but not latent forms. The data showed that MCF-7/MDR cells activate significantly more TGF-β1 than MCF-7/WT cells, although this activation was decreased by siPSG1 or DCM (Figure 3C). The effect of siPSG1 or DCM on TGF-β1 activation could be rescued by PSG1 protein at different concentrations (Figure 3D). Furthermore, the activation of TGF-β1 by PSG1 was confirmed by SMAD binding element (SBE) luciferase reporter that contains a firefly luciferase gene under the control of SBE. Addition of PSG1 to siPSG1 or DCM treated MCF-7/MDR cells restored the luciferase activity and thus indicates activation of TGF-β1 [27] (Figure 3E). Therefore, these data suggested that PSG1 can activate TGF-β1 in MCF-7/MDR cells to modulate chemoresistance. However, knocking down PSG1 protein did not change the expression of TGF-β1, probably because PSG1 activates TGF-β1 by interacting with the latent complex of TGF-β1, but not by modulating gene expression [24,25] (Supplementary Figure S4).

Although TGF-β1 activation triggers pro-proliferation signals [28], knocking down TGF-β1 diminished the chemoresistance of MCF-7/MDR cells to ADM and PTX separately (Figure 3F). Among the pro-proliferation signals, the epithelial–mesenchymal transition (EMT) signal is tightly linked to chemoresistance [29]. As expected, DCM treatment inhibited the expression of the hallmarks in the EMT pathway N-cadherin, vimentin and fibronectin, whereas addition of recombinant PSG1 restored them (Figure 3G), suggesting that DCM modulates chemoresistance via a PSG1–TGF-β1–EMT pathway.

Validation of the effect of DCM in vivo and in clinical drug-sensitivity tests

We analysed the inhibitory effect of DCM in vivo. MCF-7/MDR xenografts were generated in athymic nude mice, then the mice were treated with 25 μg/ml of DCM, along with treatment with 2.5 mg/kg ADM or PTX at the site of xenografts. DCM, ADM or PTX alone slightly decreased the volume of the xenografts. When ADM or PTX was combined with DCM, the shrinkage of xenografts increased significantly, although sera addition of PSG1 could attenuate the effect of DCM (Figure 4A).

Anti-chemoresistant effect of DCM in vivo and in clinical samples

Figure 4
Anti-chemoresistant effect of DCM in vivo and in clinical samples

(A) Left panel: DCM inhibits growth of MCF-7/MDR xenografts in athymic nude mice. The xenografts were allowed to grow up to ∼100 m3 after MCF-7/MDR inoculation (0 days after treatment), then ADM, PTX, DCM, ADM+DCM or PTX+DCM were applied for 30 days to assess the volume change in xenografts. ***P<0.001. Right panel: MCF-7/MDR xenografts bearing athymic nude mice were treated with PTX, ADM and DCM in the presence or absence of recombinant PSG1, then the volume of xenografts were measured at the 20th day of treatment. **P<0.01. (B) DCM showed cytotoxicity on primary breast cancer cells. Cell viability was analysed in CD-DST system, DCM untreated cells were used a control whose cell viability was normalized to 100%. **P<0.01. (C and D) Primary breast cancer cells were treated with different concentration of EPI (C) or PTX (D) in the presence or absence of DCM in CD-DST system, inhibition ratio (IR) of different treatment was calculated as (%)=[(AB)/A] × 100, where A is the average cell viability of primary breast cancer cells, and B is the viability of the treated cells. (E and F) Primary chemoresistant breast cancer cells were treated with different concentration of EPI (E) or PTX (F) in the presence or absence of DCM in CD-DST system, IR of different treatment was analysed.

Figure 4
Anti-chemoresistant effect of DCM in vivo and in clinical samples

(A) Left panel: DCM inhibits growth of MCF-7/MDR xenografts in athymic nude mice. The xenografts were allowed to grow up to ∼100 m3 after MCF-7/MDR inoculation (0 days after treatment), then ADM, PTX, DCM, ADM+DCM or PTX+DCM were applied for 30 days to assess the volume change in xenografts. ***P<0.001. Right panel: MCF-7/MDR xenografts bearing athymic nude mice were treated with PTX, ADM and DCM in the presence or absence of recombinant PSG1, then the volume of xenografts were measured at the 20th day of treatment. **P<0.01. (B) DCM showed cytotoxicity on primary breast cancer cells. Cell viability was analysed in CD-DST system, DCM untreated cells were used a control whose cell viability was normalized to 100%. **P<0.01. (C and D) Primary breast cancer cells were treated with different concentration of EPI (C) or PTX (D) in the presence or absence of DCM in CD-DST system, inhibition ratio (IR) of different treatment was calculated as (%)=[(AB)/A] × 100, where A is the average cell viability of primary breast cancer cells, and B is the viability of the treated cells. (E and F) Primary chemoresistant breast cancer cells were treated with different concentration of EPI (E) or PTX (F) in the presence or absence of DCM in CD-DST system, IR of different treatment was analysed.

The toxicity of DCM was also monitored in normal mice. DCM alone seldom killed the mice within 30 days of treatment, while treating it together with ADM or PTX killed a small number of mice, but the level did not reach significance (Supplementary Figure S5).

The effect of DCM was confirmed in clinical drug sensitive tests with a CD-DST system that mimics the physiological and physical environment of cancer cells in the body [10]. Primary cells from each patient (Supplementary Table S5) were treated with either epidoxorubicin (EPI)–DCM or PTX–DCM combinations. In these experiments, EPI was used instead of ADM because this was the choice of physicians in the hospital. However, both drugs are anthracyclines, so their activity would not show major differences in outcomes. In each combination of drugs, primary cells from each patient were treated with two doses of PTX or EPI.

First, all collected ER-negative primary cancer cells were treated with different combinations of EPI, PTX and DCM regardless of chemoresistant status. DCM alone triggered a 22±1.5% IR in the cancer cells (Figure 4B, n=48). EPI alone had an IR of 26.70±4.04% at 0.1 μg/ml and 43.97±5.65% at 0.3 μg/ml, which was enhanced to 39.60±4.56% and 72.17±3.53% respectively, when combined with DCM (Figure 4C). Similarly, the IRs of PTX increased from 35.53±4.28% and 62.65±4.12% to 47.14.9±3.57% and 73.31±3.19% at 10 and 20 μg/ml respectively (Figure 4D). These results suggested that DCM improves the overall effect of TA-based treatment.

Second, we separately analysed the effect of DCM on chemoresistant primary breast cancer cells, which were defined as having an IR of <30% for ADM or PTX. The IRs of 0.1 and 0.3 μg/ml EPI were 16.99±2.08% and 25.39±3.90%, and these were enhanced to 30.26±5.09% and 64.51±4.84% when combined with DCM, with 43% and 60% increases respectively (Figure 4E). Similarly, the IRs of PTX increased from 8.60±3.19% and 35.31±3.77% to 36.01±3.16% and 62.10±4.23% at 10 and 20 μg/ml, with increases of 73% and 43% respectively (Figure 4F). These results indicated that DCM is effective in re-sensitizing chemoresistant cancer cells in clinical samples.

Using all three of DCM, ADM and PTX in combination was not tested on the primary breast cancer cells, because the cell number required for CD-DST is small, and most cells were killed by the combination of ADM and PTX.

DISCUSSION

ER and progesterone receptor (PR) are two important hormone receptors that modulate breast cancer biological and pathological features. However, the clinical significance of PR is controversial especially in ER-negative breast cancers. Some studies indicate that the ER-negative/PR-positive subtype is a clinically useful entity with incidence of 1–4% [30,31], whereas others claim this subtype is caused by technical artefact [3234] and too rare to show clinical significance [35]. Indeed, we also found relatively few numbers of ER-negative/PR-positive tumours both in the public datasets and clinical samples. On the other hand, ER testing, as well as the use of ER status to guide cancer therapy, are now well established [1,36,37]. In both prospective and retrospective studies, it was common to group the breast cancer patients and choose appropriate hormonal/chemo-therapies according to ER status. Therefore, in this study, we explored the possible molecular mechanism of chemoresistance in ER-negative breast cancers irrespective of PR status.

Furthermore, we analysed the TASig regardless of HER-2 expression because the number of patients was not big enough to make a statistical comparison between different ER/HER-2 subtypes. Also the patients involved in this study did not receive any HER-2 targeting therapy, so the underlining signalling pathway of HER-2 in mediating chemoresistance could be detected from the list of significantly changed genes between chemoresistance and sensitive patients. However, because HER-2 can modulate the pathological feature and chemotherapeutic response of breast cancers [38,39], further studies with bigger sample size may be required to explore whether HER-2 expression correlates with TASig-related signals and thus modulate chemoresistance in ER-negative breast cancers.

A biological state, whether physiological or pathological, can be described in terms of a genomic signature. Similarly, the TASig may present the gene expression pattern of a large proportion of TA-resistant ER-negative breast cancer patients because genes in the signature significantly changed between chemoresistant and sensitive patients. The effectiveness of TASig in predicting chemoresistant properties in different gene expression profiling datasets lends additional supports this idea.

We then identified that PSG1 may play a central role in the TASig, and PSG1 was predictive for chemoresponse of ER-negative breast cancer patients. The predictive power of PSG1 was weaker than that of multivariate TASig, because chemoresistance is influenced by a network of different genes, and it cannot be solely determined by PSG1 even if it plays a central role. Furthermore, patients predicted to be treatment sensitive by PSG1 had a significant reduction in the risk of distant relapse after TA-based chemotherapy. However, we separately demonstrated that PSG1 was not prognostic using available data from three retrospective cohorts that did not receive chemotherapy for ER-negative breast cancers. Previous studies showed that sera PSG1 is associated with progressive features and poor prognosis [40], but this study did not record the molecular subtypes and chemotherapeutic information of tested patients. Therefore, further study is needed to evaluate the prognostic effect of PSG1 according to different tumour subtypes and treatments.

The genes in the TASig have been reported to be associated with hallmarks of cancer. Thus, disrupting their signalling pathways may interfere with cancer progression, including the development of chemoresistance. We found inhibition of PSG1 could effectively diminish chemoresistance. DCM, which modulates expression of PSG1 in chemoresistant breast cancer cells, improves the chemotherapeutic efficacy in ER-negative breast cancers.

DCM is an FDA-approved anti-coagulant that works by inhibiting vitamin K epoxide reductase. It also can inhibit activity of NAD(P)H:quinone oxidoreductase (NQO1), which results in redox cycling with generation of superoxide (O2•−) [41]. However, oxidative stress is not enough to explain the cytotoxicity of DCM, because DCM alone kills significantly smaller numbers of non-cancerous cells or chemosensitive cancer cells. DCM might target signalling pathways that are essential for chemoresistant development, so that the action of cytotoxic chemotherapeutic agents, i.e. ADM and PTX could be enhanced. We later identified that MYC was a DCM responsive transcription factor, and MYC–PSG1–TGF-β–EMT pathway was involved in the anti-chemoresistant activity of DCM. It is widely accepted that EMT pathway contributes to the development of chemoresistance [42]. Upon EMT activation, the cancer cells are more insensitive to apoptosis signals, activating more proliferative signals, and thus show chemoresistance. Therefore, DCM reduces the protective effect of EMT from chemoresistance cancer cells, and these cells become easier to be killed by chemotherapeutic drugs. What requires further investigation is how PSG1 activates TGF-β1; it may use a pathway similar to that of thrombospondin 1 to release the active TGF-β1 from its latent form after the secretion of TGF-β1 [24]. Consistently, we did not find expression change of TGF-β1 during treatment of PGS1 proteins or knocking down of PSG1, suggesting the PSG1 associates with existing TGF-β1 proteins.

The anti-chemoresistant activity of DCM in vivo and in clinical drug-sensitivity tests was impressive, further suggesting DCM could be used to enhance chemotherapeutic efficacy. There are several types of DCM capsules/tablets available, ranging from 20 to 100 mg. These doses are safe to humans as anti-coagulants when properly used. We expect standard clinical trials will be performed in future to apply DCM in cancer patients. It is noteworthy that the DCM is an anti-coagulant, so cancer patients with bleeding tendency should be monitored carefully when treating with DCM.

Taken together, the results of our study highlight PSG1 in mediating chemoresistance in ER-negative breast cancers. In turn, DCM represents a new candidate for overcoming chemoresistance by modulating PSG1.

AUTHOR CONTRIBUTION

Dong-Xu He, Jun-jun Hao, Jian Wu and Jin-ke Wang carried out the gene expression data analysis and transcription factor analysis; Xiao-Ting Gu, Ai-Qin Mao and Guang-yuan Zhang carried out the cellular experiments; Feng Gu and Li Fu participated in the clinical validation; Dong-Xu He, Feng Gu, Jun-jun Hao, Li Fu, Zhong-yang Ding and Xin Ma participated in design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.

We thank Prof. I.C. Bruce for critical reading of the manuscript.

FUNDING

This work was supported by the China National Natural Science Foundation [grant numbers 81572940, 81622007 and 91439131 (to X.M.), 31550006 (to D.-x.H.)]; the Natural Science Foundation for Distinguished Young Scholars of Jiangsu Province [grant number BK20140004 (to X.M.)]; the National High Technology Research and Development Program (863 Program) of China [grant number SQ2015AA020948 (to X.M.)]; and Fundamental Research Funds for the Central Universities [grant numbers JUSRP51311A and JUSRP51615B (to X.M. and D.-x.H.)].

Abbreviations

     
  • ADM

    adriamycin

  •  
  • CD-DST

    collagen-gel droplet embedded-culture drug-sensitivity test

  •  
  • ChIP

    chromatin immunoprecipitation assay

  •  
  • DCM

    dicumarol

  •  
  • DRFS

    distance relapse free survival

  •  
  • EMT

    epithelial–mesenchymal transition

  •  
  • EPI

    epidoxorubicin

  •  
  • ER

    oestrogen receptor

  •  
  • NPV

    negative predictive value

  •  
  • PPV

    positive predictive value

  •  
  • PR

    progesterone receptor

  •  
  • PSG1

    pregnancy specific beta-1-glycoprotein 1

  •  
  • PTX

    paclitaxel

  •  
  • SBE

    SMAD binding element

  •  
  • TA-based

    taxane-anthracycline-based

  •  
  • TASig

    signature for TA-based chemotherapy

  •  
  • TGF-β

    transforming growth factor-β

References

References
1
Berry
 
D.A.
Cirrincione
 
C.
Henderson
 
I.C.
Citron
 
M.L.
Budman
 
D.R.
Goldstein
 
L.J.
Martino
 
S.
Perez
 
E.A.
Muss
 
H.B.
Norton
 
L.
, et al 
Estrogen-receptor status and outcomes of modern chemotherapy for patients with node-positive breast cancer
JAMA
2006
, vol. 
295
 (pg. 
1658
-
1667
)
[PubMed]
2
Isakoff
 
S.J.
 
Triple-negative breast cancer: role of specific chemotherapy agents
Cancer J.
2010
, vol. 
16
 (pg. 
53
-
61
)
[PubMed]
3
Bear
 
H.D.
Anderson
 
S.
Brown
 
A.
Smith
 
R.
Mamounas
 
E.P.
Fisher
 
B.
Margolese
 
R.
Theoret
 
H.
Soran
 
A.
Wickerham
 
D.L.
, et al 
The effect on tumor response of adding sequential preoperative docetaxel to preoperative doxorubicin and cyclophosphamide: preliminary results from National Surgical Adjuvant Breast and Bowel Project Protocol B-27
J. Clin Oncol.
2003
, vol. 
21
 (pg. 
4165
-
4174
)
[PubMed]
4
Henderson
 
I.C.
Berry
 
D.A.
Demetri
 
G.D.
Cirrincione
 
C.T.
Goldstein
 
L.J.
Martino
 
S.
Ingle
 
J.N.
Cooper
 
M.R.
Hayes
 
D.F.
Tkaczuk
 
K.H.
, et al 
Improved outcomes from adding sequential Paclitaxel but not from escalating Doxorubicin dose in an adjuvant chemotherapy regimen for patients with node-positive primary breast cancer
J. Clin. Oncol.
2003
, vol. 
21
 (pg. 
976
-
983
)
[PubMed]
5
Mazouni
 
C.
Kau
 
S.W.
Frye
 
D.
Andre
 
F.
Kuerer
 
H.M.
Buchholz
 
T.A.
Symmans
 
W.F.
Anderson
 
K.
Hess
 
K.R.
Gonzalez-Angulo
 
A.M.
, et al 
Inclusion of taxanes, particularly weekly paclitaxel, in preoperative chemotherapy improves pathologic complete response rate in estrogen receptor-positive breast cancers
Ann. Oncol.
2007
, vol. 
18
 (pg. 
874
-
880
)
[PubMed]
6
Guarneri
 
V.
Broglio
 
K.
Kau
 
S.W.
Cristofanilli
 
M.
Buzdar
 
A.U.
Valero
 
V.
Buchholz
 
T.
Meric
 
F.
Middleton
 
L.
Hortobagyi
 
G.N.
Gonzalez-Angulo
 
A.M.
 
Prognostic value of pathologic complete response after primary chemotherapy in relation to hormone receptor status and other factors
J. Clin. Oncol.
2006
, vol. 
24
 (pg. 
1037
-
1044
)
[PubMed]
7
Farmer
 
P.
Bonnefoi
 
H.
Anderle
 
P.
Cameron
 
D.
Wirapati
 
P.
Becette
 
V.
André
 
S.
Piccart
 
M.
Campone
 
M.
Brain
 
E.
, et al 
A stroma-related gene signature predicts resistance to neoadjuvant chemotherapy in breast cancer
Nat. Med.
2009
, vol. 
15
 (pg. 
68
-
74
)
[PubMed]
8
Hatzis
 
C.
Pusztai
 
L.
Valero
 
V.
Booser
 
D.J.
Esserman
 
L.
Lluch
 
A.
Vidaurre
 
T.
Holmes
 
F.
Souchon
 
E.
Wang
 
H.
, et al 
A genomic predictor of response and survival following taxane-anthracycline chemotherapy for invasive breast cancer
JAMA
2011
, vol. 
305
 (pg. 
1873
-
1881
)
[PubMed]
9
Potti
 
A.
Dressman
 
H.K.
Bild
 
A.
Riedel
 
R.F.
Chan
 
G.
Sayer
 
R.
Cragun
 
J.
Cottrill
 
H.
Kelley
 
M.J.
Petersen
 
R.
, et al 
Genomic signatures to guide the use of chemotherapeutics
Nat. Med.
2006
, vol. 
12
 (pg. 
1294
-
1300
)
[PubMed]
10
Takamura
 
Y.
Kobayashi
 
H.
Taguchi
 
T.
Motomura
 
K.
Inaji
 
H.
Noguchi
 
S.
 
Prediction of chemotherapeutic response by collagen gel droplet embedded culture-drug sensitivity test in human breast cancers
Int. J. Cancer
2002
, vol. 
98
 (pg. 
450
-
455
)
[PubMed]
11
Gu
 
F.
Ma
 
Y.
Zhang
 
Z.
Zhao
 
J.
Kobayashi
 
H.
Zhang
 
L.
Fu
 
L.
 
Expression of Stat3 and Notch1 is associated with cisplatin resistance in head and neck squamous cell carcinoma
Oncol. Rep.
2010
, vol. 
23
 (pg. 
671
-
676
)
[PubMed]
12
Horak
 
C.E.
Pusztai
 
L.
Xing
 
G.
Trifan
 
O.C.
Saura
 
C.
Tseng
 
L.M.
Chan
 
S.
Welcher
 
R.
Liu
 
D.
 
Biomarker analysis of neoadjuvant doxorubicin/cyclophosphamide followed by ixabepilone or Paclitaxel in early-stage breast cancer
Clin. Cancer Res.
2013
, vol. 
19
 (pg. 
1587
-
1595
)
[PubMed]
13
Hanahan
 
D.
Weinberg
 
R.A.
 
Hallmarks of cancer: the next generation
Cell
2011
, vol. 
144
 (pg. 
646
-
674
)
[PubMed]
14
He
 
D.X.
Xia
 
Y.D.
Gu
 
X.T.
Jin
 
J.
Ma
 
X.
 
A 20-gene signature in predicting the chemoresistance of breast cancer to taxane-based chemotherapy
Mol. Biosyst.
2014
, vol. 
10
 (pg. 
3111
-
31119
)
[PubMed]
15
He
 
D.X.
Xia
 
Y.D.
Gu
 
X.T.
Jin
 
J.
Ma
 
X.
 
A transcription/translation-based gene signature predicts resistance to chemotherapy in breast cancer
J. Pharm. Biomed. Anal.
2014
, vol. 
102C
 (pg. 
500
-
508
)
16
Shannon
 
P.
Markiel
 
A.
Ozier
 
O.
Baliga
 
N.S.
Wang
 
J.T.
Ramage
 
D.
Amin
 
N.
Schwikowski
 
B.
Ideker
 
T.
 
Cytoscape: a software environment for integrated models of biomolecular interaction networks
Genome Res
2003
, vol. 
13
 (pg. 
2498
-
2504
)
[PubMed]
17
Montojo
 
J.
Zuberi
 
K.
Rodriguez
 
H.
Kazi
 
F.
Wright
 
G.
Donaldson
 
S.L.
Morris
 
Q.
Bader
 
G.D.
 
GeneMANIA Cytoscape plugin: fast gene function predictions on the desktop
Bioinformatics
2010
, vol. 
26
 (pg. 
2927
-
2928
)
[PubMed]
18
Datta
 
J.
Kutay
 
H.
Nasser
 
M.W.
Nuovo
 
G.J.
Wang
 
B.
Majumder
 
S.
Liu
 
C.G.
Volinia
 
S.
Croce
 
C.M.
Schmittgen
 
T.D.
, et al 
Methylation mediated silencing of MicroRNA-1 gene and its role in hepatocellular carcinogenesis
Cancer Res.
2008
, vol. 
68
 (pg. 
5049
-
5058
)
[PubMed]
19
Aure
 
M.R.
Leivonen
 
S.K.
Fleischer
 
T.
Zhu
 
Q.
Overgaard
 
J.
Alsner
 
J.
Tramm
 
T.
Louhimo
 
R.
Alnæs
 
G.I.
Perälä
 
M.
, et al 
Individual and combined effects of DNA methylation and copy number alterations on miRNA expression in breast tumors
Genome Biol.
2013
, vol. 
14
 pg. 
R126
 
[PubMed]
20
Buffa
 
F.M.
Camps
 
C.
Winchester
 
L.
Snell
 
C.E.
Gee
 
H.E.
Sheldon
 
H.
Taylor
 
M.
Harris
 
A.L.
Ragoussis
 
J.
 
microRNA-associated progression pathways and potential therapeutic targets identified by integrated mRNA and microRNA expression profiling in breast cancer
Cancer Res.
2011
, vol. 
71
 (pg. 
5635
-
5645
)
[PubMed]
21
Kuo
 
W.H.
Chang
 
Y.Y.
Lai
 
L.C.
Tsai
 
M.H.
Hsiao
 
C.K.
Chang
 
K.J.
Chuang
 
E.Y.
 
Molecular characteristics and metastasis predictor genes of triple-negative breast cancer: a clinical study of triple-negative breast carcinomas
PLoS One
2012
, vol. 
7
 pg. 
e45831
 
[PubMed]
22
Esserman
 
L.J.
Berry
 
D.A.
Cheang
 
M.C.
Yau
 
C.
Perou
 
C.M.
Carey
 
L.
DeMichele
 
A.
Gray
 
J.W.
Conway-Dorsey
 
K.
Lenburg
 
M.E.
, et al 
Chemotherapy response and recurrence-free survival in neoadjuvant breast cancer depends on biomarker profiles: results from the I-SPY 1 TRIAL (CALGB 150007/150012; ACRIN 6657)
Breast Cancer Res. Treat.
2012
, vol. 
132
 (pg. 
1049
-
1062
)
[PubMed]
23
Ma
 
X.
Cai
 
Y.
He
 
D.
Zou
 
C.
Zhang
 
P.
Lo
 
C.Y.
Xu
 
Z.
Chan
 
F.L.
Yu
 
S.
Chen
 
Y.
, et al 
Transient receptor potential channel TRPC5 is essential for P-glycoprotein induction in drug-resistant cancer cells
Proc. Natl. Acad. Sci. U.S.A.
2012
, vol. 
109
 (pg. 
16282
-
16287
)
[PubMed]
24
Blois
 
S.M.
Sulkowski
 
G.
Tirado-Gonzalez
 
I.
Warren
 
J.
Freitag
 
N.
Klapp
 
B.F.
Rifkin
 
D.
Fuss
 
I.
Strober
 
W.
Dveksler
 
G.S.
 
Pregnancy-specific glycoprotein 1 (PSG1) activates TGF-beta and prevents dextran sodium sulfate (DSS)-induced colitis in mice
Mucosal. Immunol.
2014
, vol. 
7
 (pg. 
348
-
358
)
[PubMed]
25
Snyder
 
S.K.
Wessner
 
D.H.
Wessells
 
J.L.
Waterhouse
 
R.M.
Wahl
 
L.M.
Zimmermann
 
W.
Dveksler
 
G.S.
 
Pregnancy-specific glycoproteins function as immunomodulators by inducing secretion of IL-10, IL-6 and TGF-beta1 by human monocytes
Am. J. Reprod. Immunol.
2001
, vol. 
45
 (pg. 
205
-
216
)
[PubMed]
26
Bhola
 
N.E.
Balko
 
J.M.
Dugger
 
T.C.
Kuba
 
M.G.
Sanchez
 
V.
Sanders
 
M.
Stanford
 
J.
Cook
 
R.S.
Arteaga
 
C.L.
 
TGF-beta inhibition enhances chemotherapy action against triple-negative breast cancer
J. Clin. Invest.
2013
, vol. 
123
 (pg. 
1348
-
1358
)
[PubMed]
27
Moustakas
 
A.
Souchelnytskyi
 
S.
Heldin
 
C.H.
 
Smad regulation in TGF-beta signal transduction
J. Cell Sci.
2001
, vol. 
114
 
Pt 24
(pg. 
4359
-
4369
)
[PubMed]
28
Derynck
 
R.
Zhang
 
Y.E.
 
Smad-dependent and Smad-independent pathways in TGF-beta family signalling
Nature
2003
, vol. 
425
 (pg. 
577
-
584
)
[PubMed]
29
Roberts
 
A.B.
Tian
 
F.
Byfield
 
S.D.
Stuelten
 
C.
Ooshima
 
A.
Saika
 
S.
Flanders
 
K.C.
 
Smad3 is key to TGF-beta-mediated epithelial-to-mesenchymal transition, fibrosis, tumor suppression and metastasis
Cytokine Growth Factor Rev
2006
, vol. 
17
 (pg. 
19
-
27
)
[PubMed]
30
Rhodes
 
A.
Jasani
 
B.
 
The oestrogen receptor-negative/progesterone receptor-positive breast tumour: a biological entity or a technical artefact?
J. Clin. Pathol.
2009
, vol. 
62
 (pg. 
95
-
96
)
[PubMed]
31
Shen
 
T.
Brandwein-Gensler
 
M.
Hameed
 
O.
Siegal
 
G.P.
Wei
 
S.
 
Characterization of estrogen receptor-negative/progesterone receptor-positive breast cancer
Hum. Pathol.
2015
, vol. 
46
 (pg. 
1776
-
1784
)
[PubMed]
32
De Maeyer
 
L.
Van Limbergen
 
E.
De Nys
 
K.
Moerman
 
P.
Pochet
 
N.
Hendrickx
 
W.
Wildiers
 
H.
Paridaens
 
R.
Smeets
 
A.
Christiaens
 
M.R.
, et al 
Does estrogen receptor negative/progesterone receptor positive breast carcinoma exist?
J. Clin. Oncol.
2008
, vol. 
26
 (pg. 
335
-
356
)
33
Nadji
 
M.
Gomez-Fernandez
 
C.
Ganjei-Azar
 
P.
Morales
 
A.R.
 
Immunohistochemistry of estrogen and progesterone receptors reconsidered: experience with 5,993 breast cancers
Am. J. Clin. Pathol.
2005
, vol. 
123
 (pg. 
21
-
27
)
[PubMed]
34
Hefti
 
M.M.
Hu
 
R.
Knoblauch
 
N.W.
Collins
 
L.C.
Haibe-Kains
 
B.
Tamimi
 
R.M.
Beck
 
A.H.
 
Estrogen receptor negative/progesterone receptor positive breast cancer is not a reproducible subtype
Breast Cancer Res.
2013
, vol. 
15
 pg. 
R68
 
[PubMed]
35
Olivotto
 
I.A.
Truong
 
P.T.
Speers
 
C.H.
Bernstein
 
V.
Allan
 
S.J.
Kelly
 
S.J.
Lesperance
 
M.L.
 
Time to stop progesterone receptor testing in breast cancer management
J. Clin. Oncol.
2004
, vol. 
22
 (pg. 
1769
-
1770
)
[PubMed]
36
Bartlett
 
J.M.
Ellis
 
I.O.
Dowsett
 
M.
Mallon
 
E.A.
Cameron
 
D.A.
Johnston
 
S.
Hall
 
E.
A'Hern
 
R.
Peckitt
 
C.
Bliss
 
J.M.
, et al 
Human epidermal growth factor receptor 2 status correlates with lymph node involvement in patients with estrogen receptor (ER) negative, but with grade in those with ER-positive early-stage breast cancer suitable for cytotoxic chemotherapy
J. Clin. Oncol.
2007
, vol. 
25
 (pg. 
4423
-
4430
)
[PubMed]
37
Bernsdorf
 
M.
Ingvar
 
C.
Jorgensen
 
L.
Tuxen
 
M.K.
Jakobsen
 
E.H.
Saetersdal
 
A.
Kimper-Karl
 
M.L.
Kroman
 
N.
Balslev
 
E.
Ejlertsen
 
B.
 
Effect of adding gefitinib to neoadjuvant chemotherapy in estrogen receptor negative early breast cancer in a randomized phase II trial
Breast Cancer Res. Treat.
2011
, vol. 
126
 (pg. 
463
-
470
)
[PubMed]
38
Ahmed
 
K.M.
Cao
 
N.
Li
 
J.J.
 
HER-2 and NF-kappaB as the targets for therapy-resistant breast cancer
Anticancer Res.
2006
, vol. 
26
 (pg. 
4235
-
4243
)
[PubMed]
39
Zhou
 
B.P.
Liao
 
Y.
Xia
 
W.
Zou
 
Y.
Spohn
 
B.
Hung
 
M.C.
 
HER-2/neu induces p53 ubiquitination via Akt-mediated MDM2 phosphorylation
Nat. Cell Biol.
2001
, vol. 
3
 (pg. 
973
-
982
)
[PubMed]
40
Kuhajda
 
F.P.
Bohn
 
H.
Mendelsohn
 
G.
 
Pregnancy-specific beta-1 glycoprotein (SP-1) in breast carcinoma. Pathologic and clinical considerations
Cancer
1984
, vol. 
54
 (pg. 
1392
-
1396
)
[PubMed]
41
Dinkova-Kostova
 
A.T.
Talalay
 
P.
 
Persuasive evidence that quinone reductase type 1 (DT diaphorase) protects cells against the toxicity of electrophiles and reactive forms of oxygen
Free Radic. Biol. Med.
2000
, vol. 
29
 (pg. 
231
-
240
)
[PubMed]
42
Mani
 
S.A.
Guo
 
W.
Liao
 
M.J.
Eaton
 
E.N.
Ayyanan
 
A.
Zhou
 
A.Y.
Brooks
 
M.
Reinhard
 
F.
Zhang
 
C.C.
Shipitsin
 
M.
, et al 
The epithelial-mesenchymal transition generates cells with properties of stem cells
Cell
2008
, vol. 
133
 (pg. 
704
-
715
)
[PubMed]

Author notes

1

These authors contributed equally to this study.

Supplementary data