PDGFRβ is an essential therapeutic target for BRCA1-deficient mammary tumors

Basal-like breast cancers (BLBCs) are a leading cause of cancer death due to their capacity to metastasize and lack of effective therapies. More than half of BLBCs have a dysfunctional BRCA1. Although most BRCA1-deficient cancers respond to DNA-damaging agents, resistance and tumor recurrence remain a challenge to survival outcomes for BLBC patients. Additional therapies targeting the pathways aberrantly activated by BRCA1 deficiency are urgently needed. Most BRCA1-deficient BLBCs carry a dysfunctional INK4-RB pathway. Thus, we created genetically engineered mice with Brca1 loss and deletion of p16INK4A, or separately p18INK4C, to model the deficient INK4-RB signaling in human BLBC. By using these mutant mice and human BRCA1-deficient and proficient breast cancer tissues and cells, we tested if there exists a druggable target in BRCA1-deficient breast cancers. Heterozygous germline or epithelium-specific deletion of Brca1 in p18INK4C- or p16INK4A-deficient mice activated Pdgfrβ signaling, induced epithelial-to-mesenchymal transition, and led to BLBCs. Confirming this role, targeted deletion of Pdgfrβ in Brca1-deficient tumor cells promoted cell death, induced mesenchymal-to-epithelial transition, and suppressed tumorigenesis. Importantly, we also found that pharmaceutical inhibition of Pdgfrβ and its downstream target Pkcα suppressed Brca1-deficient tumor initiation and progression and effectively killed BRCA1-deficient cancer cells. Our work offers the first genetic and biochemical evidence that PDGFRβ-PKCα signaling is repressed by BRCA1, which establishes PDGFRβ-PKCα signaling as a therapeutic target for BRCA1-deficient breast cancers.


Background
The heterogeneity of breast cancer is marked by pathologically distinct tumor types that differ in their responsiveness to treatment. In the clinical setting, breast cancer is comprised of three main subtypes: HER2positive, estrogen receptor (ER)-positive luminal, and basal-like cancers (BLBCs) that make up the majority of triple-negative breast cancer [TNBC; ER, PR, HER2 negative by immunohistochemistry (IHC) analysis] [1,2]. BLBCs are highly heterogeneous and aggressive, perhaps due to their enrichment of tumor-initiating cells (TICs) or cancer stem cells that are thought to drive clinical relapse and metastasis [3][4][5][6]. TICs and tumor cells with mesenchymal features have enhanced capacity to metastasize and are resistant to radio-and chemotherapy [7]. Mammary TICs can be generated from luminal tumor cells by an epithelial-to-mesenchymal transition (EMT) program [7][8][9][10][11][12][13][14][15], a process in which epithelial cells lose many of their epithelial characteristics and acquire mesenchymal features [7]. BLBCs likely originate from luminal progenitors [8][9][10][11] and contain a number of distinct cell types including cells that express luminal biomarkers [16][17][18]. Notably, more than half of BLBCs are associated with functional loss of BRCA1, caused by germline or somatic mutation or by promoter hypermethylation [19][20][21][22]. BRCA1 is a tumor suppressor that functions in DNA damage repair. Although the majority of BRCA1-deficient cancer patients respond to DNA-damaging agents such as cisplatin and poly (ADPribose) polymerase (PARP) inhibitors, tumor recurrence and resistance, likely driven by TICs, combine to decrease the 5-year survival of such patients [23,24]. Thus, additional therapies targeting the pathways aberrantly activated by BRCA1 deficiency are urgently needed.
PDGFRβ, a receptor for members of the plateletderived growth factors (PDGFs), is abundantly expressed in stromal fibroblasts [35][36][37]. PDGFRβ signaling plays a critical role in activating cancerassociated fibroblasts (CAFs) which facilitate breast cancer growth and progression [35,38,39]. Notably, PDGFRβ expression is specifically upregulated in late-stage breast cancer cells [40][41][42]. PDGFRβ activates PKCα which then phosphorylates FRA1, a key EMT-inducing transcription factor (EMT-TF), and drives TIC function of transformed MECs [43,44]. Inhibition of PDGFRβ-PKCα signaling specifically targets oncogene-transformed MECs that have undergone EMT and induces mesenchymal-toepithelial transition (MET), suppressing tumor initiation and progression [40,44,45]. It remains elusive how PDGFRβ signaling is involved in epithelial tumorigenesis. For example, whether PDGFRβ signaling activates EMT in a tumor cell autonomous manner and whether PDGFRβ-PKCα signaling is required for survival and for maintaining mesenchymal status of the tumor cells in vivo are key unanswered questions that may inform treatment strategies for breast cancer.
In this study, we examined the molecular outcomes of BRCA1 deficiency combined with loss of either p16 or p18 to identify targetable pathways in BRCA1/INK4-RBdeficient tumors. This identified activation of the Pdgfrβ pathway as potential driver in these tumors. Further, we showed preclinical evidence that these molecular alterations in Pdgfrβ pathway could be effectively therapeutically targeted.

Mammary tumor cell preparation and tumorsphere formation assay
Mammary tumors were dissected and tumor cell suspensions were prepared as previously described [11,28,31]. For the primary and secondary tumorsphere formation assays, primary p18 −/− ;Brca1 MGKO mammary tumor cells were plated onto six-well, ultra-low attachment plates, in serum-free DMEM-F12 supplemented with B27, EGF, and bFGF as described [28,31]. Primary tumorspheres formed were collected and dissociated after 9 days of culture. 10 4 cells dissociated from the primary tumorsphere were plated in triplicate with or without treatment. Secondary tumorspheres that formed after 6 days of culture were counted under the microscope.

Transplantation and tumor treatment
For in vivo transplantation, p18 −/− ;Brca1 MGKO primary tumor cells transfected with the Pdgfrβ CRISPR/Cas9 KO plasmids were FACS sorted. Six thousand GFP neg and GFP pos live cells (trypan blue negative) were suspended in a 50% solution of Matrigel (BD) and then inoculated into the left and right inguinal mammary fat pads (MFPs) of 6-week-old female NSG mice (Jackson Laboratory), respectively. Four weeks after transplantation, animals were euthanized and mammary tumors were dissected for histopathological and immunohistochemical analyses. For ex vivo transplantation, primary p18 −/− ;Brca1 MGKO tumor cells were cultured to generate primary tumorsphere. 10 4 cells dissociated from primary tumorspheres were treated with DMSO, Ro-31-8220, or PDGFR inh III for 6 days. One thousand live cells were transplanted into MFPs of NSG mice. Four weeks after transplantation, animals were euthanized and mammary tumors were analyzed. For treatment of established mammary tumors, p18 −/− ;Brca1 MGKO tumor cells were transplanted into MFPs of NSG mice and allowed to reach 200 mm 3 in size. Mice were then treated with daily i.p. injection of the agents, and the tumor size was determined.
Microarray analysis, western blot, and ChIP assay RNA was extracted and purified from tumors using an RNeasy kit (Qiagen). Tumor RNA was reverse transcribed, amplified, and labeled with Cy5, and WT mammary tissue reference RNA was reverse transcribed, amplified, and labeled with Cy3. The amplified sample and reference were co-hybridized to Agilent 4x180k custom mouse microarrays and were analyzed as described previously [14,53]. Murine tumor gene expression data was deposited at the Gene Expression Omnibus under accession number GSE155239. Single sample gene set enrichment analysis (ssGSEA) was performed using GenePattern [54]. The gene signature for PDGF signaling was taken from Reactome [55], while EMT and mammary stem cell signatures were published [56,57]. For the western blot, tissue and cell lysates were prepared as previously reported [28,31]. The primary antibodies used were BRCA1 (Santa Cruz), E-cad, Vim, PDGFRβ, p-PDGFRβ, p-PKCα, p-Fra1 and Snail (Cell signaling), Gapdh (Ambion), and Twist (Abcam). ChIP assays were carried out as previously described [28,31]. Briefly, T47D and HCC1937 cells were treated with 1.5% formaldehyde and sonicated. Anti-BRCA1 antibody (D-9, Santa Cruz) or control mouse IgG was used to precipitate chromatin associated with BRCA1. Q-PCR was performed to determine the relative abundance of target DNA. Specific primers for the analysis of BRCA1 binding to PDGFRβ are listed in Additional file 1.

Human tumor samples and gene expression datasets
Formalin-fixed paraffin-embedded (FFPE) human breast cancer samples lacking patient-identifying information were obtained with IRB approval from the Tissue Banks at the University of Miami and the Department of Pathology at Shenzhen University. All samples obtained were non-treated invasive breast cancers with known ER status. The MetaBric human breast cancer dataset [58] was analyzed for correlation between BRCA1 with PDGFRβ and PKCα mRNA levels.

Statistical analysis
All data are presented as the mean ± SD for at least three repeated individual experiments for each group. Quantitative results were analyzed by the two-tailed Fisher exact test or two-tailed Student's t test. p < 0.05 was considered statistically significant.
Implying epithelial origin, most Pdgfrβ-positive p18 −/− ; Brca1 MGKO and p16 −/− ;Brca1 MGKO mammary tumor cells were also positive for Ck14, an epithelial marker. Notably, Pdgfrβ and Ck14 doubly positive tumor cells were also detected in lung metastases of p18 −/− ; Brca1 MGKO and p16 −/− ;Brca1 MGKO mammary tumors ( Fig. 2a, b, and Additional file 3), which demonstrated that the metastases were derived from Brca1-deficient basal-like mammary tumors with an increased Pdgfrβ expression. Using western blots to assess protein, we found that the expression of Pdgfrβ, p-Pkcα, and p-Fra1 was clearly increased in p18 −/− ;Brca1 MGKO tumors relative to that of adjacent tumor-free mammary tissues in the same mice or to p18 −/− tumors (Fig. 2d, Additional file 4), and in p18 −/− ;Brca1 MGKO tumor cells relative to that in p18 −/− tumor cells (Fig. 2e). As a whole, these data suggest that loss of Brca1 induces EMT and metastatic basal-like tumors in an epitheliumautonomous manner and that Brca1 loss activates Pdgfrβ-Pkcα signaling in epithelial tumor cells thereby enriching the number of tumor cells with EMT-like features.

BRCA1 binds to the PDGFRβ locus to repress its transcription
We and others have shown that BRCA1 binds to FOXC1/2 and TWIST loci to repress their transcription [31,60]. Our bioinformatic analysis revealed that there exist at least four putative BRCA1 binding sites in the PDGFRβ locus (Fig. 3c). We performed chromatinimmunoprecipitation (ChIP) assays and found that endogenous and exogenous BRCA1 bound to the PDGFRβ locus in T47D and HCC1937 cells, respectively (Fig. 3d, e). Together with the inhibitory effect of BRCA1 for the expression of PDGFRβ, these data suggest that BRCA1 binds to the PDGFRβ locus to repress its transcription.

Targeted deletion of Pdgfrβ in Brca1-deficient tumor cells promotes cell death and MET suppressing tumorigenesis
To determine the role of Pdgrfβ in Brca1-deficient tumorigenesis, we knocked out Pdgfrβ in p18 −/− ; Brca1 MGKO primary tumor cells by using the CRISPR/ Cas9 system. We transfected Pdgfrβ CRISPR/Cas9 KO plasmids (encoding GFP) into p18 −/− ;Brca1 MGKO primary tumor cells. Two days after transfection, GFPnegative (GFP neg ) and GFP-positive (GFP pos ) cells were FACS sorted for further analysis. We found that GFP pos tumor cells expressed nearly no detectable Pdgfrβ and drastically reduced p-Pkcα, p-Fra1, and Vim, when compared with GFP neg tumor cells (Fig. 4a). These data confirmed successful depletion of Pdgfrβ protein in the GFP pos p18 −/− ;Brca1 MGKO mammary tumor cells. Additionally, these data also indicate that activation of Pkcα and its downstream target Fra1 is dependent on Pdgfrβ signaling, and Pdgfrβ signaling is required for maintaining the mesenchymal traits in Brca1-deficient tumor cells.
To test the impact of Pdgfrβ CRISPR/Cas9 KO on tumor cell viability, we performed annexin V staining. We found that 8.8% of GFP pos (Pdgfrβ KO) tumor cells were positive for annexin V compared to 4.3% of the GFP neg (Pdgfrβ WT) cells (Fig. 4b). These data indicate that acute deletion of Pdgfrβ slightly enhances apoptosis of Brca1-deficient tumor cells. We further analyzed FACS-sorted Pdgfrβ WT and Pdgfrβ KO p18 −/− ; Brca1 MGKO tumor cells in dishes and observed that the population with characteristic morphological features of apoptosis (i.e., cell shrinkage, pyknosis, dense cytoplasm with tightly packed organelles) or dead cells was significantly increased in Pdgfrβ KO cells relative to that in Pdgfrβ WT cells after 5-7 days of culture (Fig. 4c). We found that the viable cells were significantly less and the dead cells were drastically more in Pdgfrβ KO population than in Pdgfrβ WT population (Fig. 4d, e). These data indicate that long-term depletion of Pdgfrβ in p18 −/− ;Brca1 MGKO tumor cells in vitro increases tumor cell death.
We next transplanted freshly FACS-sorted Pdgfrβ WT and Pdgfrβ KO p18 −/− ;Brca1 MGKO tumor cells that were viable into MFPs of NSG mice. We found that Pdgfrβ WT p18 −/− ;Brca1 MGKO tumor cells generated palpable tumors in 10-14 days whereas Pdgfrβ KO cells did not generate palpable tumors in the same time period. Four weeks after transplantation, tumors generated by Pdgfrβ KO cells were significantly smaller than tumors generated by Pdgfrβ WT cells (Fig. 4f). We performed western blot in conjunction with IHC analysis and observed that tumors generated by Pdgfrβ KO cells expressed clearly reduced levels of Pdgfrβ, p-Pkcα, p-Fra1, Vim, Snail, and Twist, but increased levels of E-cad, when compared with tumors generated by Pdgfrβ WT cells (Fig. 4g, h, and data not shown). Notably, we detected that some cells in tumors generated by Pdgfrβ KO p18 −/− ; Brca1 MGKO cells expressed high levels of E-cad, whereas E-cad-positive cells in tumors generated by Pdgfrβ WT p18 −/− ;Brca1 MGKO tumor cells were rarely observed (Fig.  4h). These results demonstrate that deletion of Pdgfrβ in Brca1-deficient tumor cells promotes MET and suppresses tumorigenesis.

Inhibition of Pdgfrβ or Pkcα activity suppresses Brca1deficient tumor initiating potential
To determine whether Pdgfrβ and Pkcα activity represented therapeutic target in BRCA1-deficient tumors, we treated p18 −/− ;Brca1 MGKO tumor cells with a PDGFR Inh III and PKCα inhibitor, Ro-31-8220. The activity and specificity of these drugs in inhibiting phosphorylation of PDGFRβ and PKCα have been well confirmed by multiple groups [44,51,52,61]. Illustrating these inhibitors were having the intended The results were normalized to the amount of input and compared with the IgG-negative controls. Data are represented as mean ± SD. e ChIP analysis of exogenous BRCA1 binding to the PDGFRβ locus in HCC1937 cells transfected with Babe-empty (Empty) or pBabe-HA-BRCA1 (BRCA1). The results were normalized to the amount of Input and compared with the empty controls. Data are represented as mean ± SD effects, we observed reduced Pdgfrβ and Pkcα phosphorylation following treatment (Fig. 5a, and Additional file 6A, B). In addition, we examined the impact of dosage on tumor cells. Treatment of p18 −/− ;Brca1 MGKO tumor cells with PDGFR Inh III and Ro-31-8220 significantly reduced cell number and increased cell death, particularly with high dosages (200 nM for PDGFR Inh III and 350 nM for Ro-31-8220) (Fig. 5b, c). At a low dosage (20 nM for PDGFR Inh III and 35 nM for Ro-31-8220), these drugs converted spindle-shaped, mesenchymal-like cells into epithelial-like cells (Fig. 5d). This is consistent with the data derived from targeted deletion of Pdgfrβ, and these results further confirm that inhibition of We have previously reported that deletion of Brca1 in p18-deficient mice activates EMT and enhances TICs [31]. We detected that p18 −/− ;Brca1 MGKO tumor cells formed significantly more and larger spheres than p18 −/− cells (data not shown) confirming the role of loss-of-function of Brca1 in stimulating TICs. We treated p18 −/− ;Brca1 MGKO tumorsphere-dissociated cells with PDGFR Inh III or Ro-31-8220 and found that inhibition of Pdgfrβ or Pkcα activity significantly reduced secondary tumorsphere forming potential (Fig. 5e, g). Moreover, we transplanted p18 −/− ;Brca1 MGKO tumorsphere-dissociated cells into NSG mice and found that cells pretreated with low-dose PDGFR Inh III or Ro-31-8220 produced significantly smaller tumors than control cells (Fig. 5f, h). Western blot and IHC analysis revealed that tumors generated by PDGFR Inh III or Ro-31-8220 pretreated cells expressed higher levels of E-cad and lower levels of Vim (Fig. 5i, j) than DMSOpretreated cells. Again, these results confirm that the inhibition of Pdgfrβ and Pkcα activity promotes MET and reduces Brca1-deficient tumor initiating potential. 200 mm 3 in size and then mice were treated with DMSO or inhibitors daily. Three days after treatment, tumors from PDGFR Inh III-or Ro-31-8220-treated mice began to show a significant size reduction in comparison with the tumors from DMSO-treated animals (Fig. 6a, b). After a 9-day treatment, tumors from DMSO-treated mice reached 578 ± 265 mm 3 in size, whereas those from PDGFR Inh III-treated mice only reached 131 ± 22 mm 3 , which was even smaller than that of the tumor size at the start of treatment (194 ± 48 mm 3 ) (Fig. 6a), indicating that treatment with PDGFR Inh III induced regression of Brca1-deficient tumors. Consistently, animals that received Ro-31-8220 treatment for 9 days had a significant reduction in tumor size compared with DMSO treatment controls (772 ± 366 mm 3 vs. 313 ± 182 mm 3 , Fig. 6b). These data demonstrate that pharmaceutical inhibition of Pdgfrβ and Pkcα activity has therapeutic effects on established Brca1-deficient tumors.

Inhibition of Pdgfrβ or Pkcα activity suppresses established Brca1-deficient tumor progression
To determine the mechanisms associated with inhibition of tumor progression by Pdgfrβ and Pkcα inhibitors, we analyzed Brca1-deficient tumors by western blot. This revealed that all tumors treated with PDGFR Inh III or Ro-31-8220 expressed more cleaved caspase 3 and less Vim (Fig. 6c, and data not shown), and one expressed higher levels of E-cad when compared with those with DMSO treatment (tumor 5 in Fig. 6c, and data not shown). IHC analysis confirmed that all tumors treated with PDGFR Inh III or Ro-31-8220 displayed more cleaved caspase 3-positive cells than control tumors, though the level of increase varied among tumors a c d b Fig. 6 Inhibition of Pdgfrβ or Pkcα activity reduces EMT and promotes apoptosis suppressing established Brca1-deficient tumor progression. a, b p18 −/− ;Brca1 MGKO tumor cells were transplanted into MFPs of NSG mice and allowed to reach~200 mm 3 in size. Mice were then treated with daily i.p. injection of DMSO or PDGFR Inh III at 5 mg/kg (a) and DMSO or Ro-31-8220 at 5 mg/kg (b), respectively. The tumor size was determined and plotted. Data are represented as mean ± SD of four (a) and five (b) tumors in each group. *p < 0.05 between two groups at each time point by Student t test. **p < 0.05 between the group that was at the start of treatment and the group that was treated with PDGFR Inh III for 9 days. c Representative tumors treated with DMSO (three tumors) or PDGFR Inh III (three tumors) for 9 days were analyzed by western blot (left). The indicated protein levels of each lane were quantified and normalized by that of Gapdh (right). Note the increased cleaved caspase 3 and decreased Vim in PDGFR Inh III-treated samples (lanes 4, 5, and 6) in comparison with those in DMSO-treated samples (lanes 1, 2, and 3). d Representative tumors treated with DMSO or PDGFR Inh III for 9 days were analyzed by IHC. The insets show the enlarged cells that are representative and different areas of an individual tumor. All control tumors widely expressed high levels of Vim while PDGFR Inh-or Ro-31-8220-treated tumors showed faint or undetectable Vim (Fig. 6d, and data not shown). In sum, these results suggest that inhibition of Pdgfrβ or Pkcα activity suppresses tumor progression by reducing mesenchymal features and inducing apoptosis.

Inhibition of PDGFR or PKCα activity efficiently kills BRCA1-deficient human breast cancer cells
To confirm the cytotoxic effects of inhibition of PDGFR and PKCα on BRCA1-deficient human breast cancer cells, we screened six widely used cell lines including two BRCA1 WT luminal lines, T47D and MCF7; two BRCA1 WT basal lines, MDA-MB231 and BT20; and two BRCA1 mutated lines, HCC1937 and SUM149. Importantly, all these cell lines carry defective INK4-RB pathway (undetectable p16 in T47D, MCF7, MDA-MB231, BT20, SUM149, and undetectable RB in HCC1937) [62][63][64], which are similar with our p18 −/− , p18 −/− ;Brca1 MGKO , and p16 −/− ;Brca1 MGKO mouse tumor cell system. We observed that the expression of PDGFRβ and PKCα was drastically increased in two BRCA1 mutant cell lines (HCC1937 and SUM149) and a BRCA1 lowly expressing MDA-MB231 line, whereas PDGFRβ and PKCα levels were extremely low in BRCA1 WT luminal lines (T47D and MCF7) and slightly increased in a BRCA1 WT basal line (BT20) (Fig. 7a, b). Western blot analysis revealed that BRCA1 mutant HCC1937 cells expressed significantly higher amounts of PDGFRβ, p-PKCα, and p-FRA1 than BRCA1-WT T47D cells (Fig. 7b) which suggested the activation of PDGFRβ-PKCα signaling by BRCA1 loss, consistent with our finding in mice and the findings from others [44,65]. We confirmed that PDGFR inh III and Ro-31-8220 suppressed PDGFRβ and PKCα activities in HCC1937 and T47D cells, though the latter of which exhibited extremely low level of PDGFRβ and PKCα activities (Additional file 6A, B). We treated these cells with the inhibitors and found that PDGFR Inh III and Ro-31-8220 significantly reduced the number of viable cells and promoted cell death in BRCA1-mutated HCC1937 and SUM149 cells. In BRCA1 lowly expressing MDA-MB231 cells, these effects were moderate and even more slight in BRCA1 WT luminal cells (Fig. 7c, d, Additional file 6C). Importantly, BRCA1 in BT20 cells was lower than that in luminal cells (MCF7 and T47D), but not as low as that in MDA-MB231 cells. Thus, the slightly increased level of PDGFRβ and PKCα in BT20 cells relative to that in luminal cells suggested that other factors were involved in inactivation of PDGFRβ and PKCα and may explain the poor response to PDGFRβ and PKCα inhibitors in BT20 cells. Together, these data suggest that inhibition of PDGFR or PKCα activity efficiently kills BRCA1-deficient human breast cancer cells.

Expression of BRCA1 is inversely related to that of PDGF Rβ and PKCα in breast cancers
To extend these results to human breast tumors, we performed immunostaining analysis of 8 ER + and 10 ER − primary human breast cancers. We found that PDGFRβ, PKCα, FRA1, and VIM were readily detected in ER − and BRCA1 weak or non-detectable tumor cells, whereas these proteins were barely detectable in ER + and BRCA1 + tumor cells (Fig. 7e and Additional file 7A, B). Further analysis revealed that PDGFRβ and PKCα detected by immunostaining were significantly inversely correlated with BRCA1 expression (Fig. 7f). We then analyzed PDGFRβ and PKCα mRNA expression in TCGA breast cancer dataset and did not find significant differences of the expression between BRCA1 mutant and WT TNBCs (data not shown). This may be partly resulted from the small sample size of BRCA1 mutant TNBCs in TCGA dataset (PDGFRβ and PKCα mRNA data are available in only five BRCA1 mutant TNBCs). Moreover, we queried the mRNA expression of genes in the METABRIC 1584 breast cancer sample sets and found a significant inverse correlation between BRCA1 with PDGFRβ and PKCα mRNA levels (Fig. 7g). These clinical findings, consistent with our results in mice, further confirm that BRCA1 suppresses the PDGFRβ-PKCα signaling pathway in breast basal-like cancer development and progression.

Discussion
In this study, we showed that heterozygous germline deletion of or epithelium-specific deletion of Brca1 in p18deficient or p16-deficient mice led to basal-like tumors with elevated markers of EMT and Pdgfrβ-Pkcα signaling activity. We demonstrated that BRCA1 bound to the PDGFRβ locus, repressed PDGFRβ transcription, and inhibited EMT. Targeted deletion of Pdgfrβ in Brca1deficient tumor cells inactivated Pdgfrβ-Pkcα signaling, promoted cell death, induced MET, and suppressed tumorigenesis. Consistently, pharmaceutical inhibition of Pdgfrβ or Pkcα activity suppressed Brca1-deficient tumor initiation and progression. We also found that inhibition of PDGFR or PKCα activity efficiently killed BRCA1-deficient human breast cancer cells and that expression of BRCA1 was inversely related to that of PDGFRβ and PKCα in human breast cancer samples. These data not only confirm that BRCA1 suppresses EMT and basal-like tumorigenesis in an epitheliumautonomous manner, but also suggest that BRCA1 a c d e f g b Fig. 7 Inhibition of PDGFRβ or PKCα activity targets BRCA1-deficient breast cancer cells, and expression of BRCA1 is inversely related with that of PDGFRβ and PKCα in breast cancers. a Expression of BRCA1, PDGFRβ, and PKCα mRNA in human breast cancer cell lines was determined by qRT-PCR. b Expression of the proteins indicated in HCC1937 and T47D cells was determined by western blot. c, d Cancer cells were treated with DMSO, PDGFR Inh III (20 nM), or Ro-31-8220 (35 nM) and the number of viable cells (c) and dead cells (d) was determined after 5 days. Data are represented as mean ± SD of triplicates. The asterisk (*) denotes a significance from DMSO-and drug-treated cells. e Representative immunostaining analysis of serial human breast cancer sections. f Summary of the immunostaining results for human breast cancer samples. * "High" expression represents the samples positively stained with an antibody in more than 2% cells (i.e., scores equal to or higher than "+" in Additional file 7B). "Low" expression represents the samples negative or positively stained with an antibody in less than 2% cells (i.e., scores less than "+/−" in Additional file 7B). The asterisk (*) denotes a significance from BRCA1 high and BRCA1 low tumors by a two-tailed Fisher's exact test. g Correlation analysis of mRNA levels of BRCA1 and PDGFRβ or PKCα for MetaBric breast cancer patients suppresses EMT in tumor cells by repressing PDGFRβ-PKCα signaling. PDGFRβ is abundantly expressed in normal stromal fibroblasts and in late-stage breast cancer cells, whereas PDGFs, ligands to PDGFRβ, are mainly expressed and secreted in epithelial and carcinoma cells [35][36][37]40]. Importantly, in human breast cancers, high stromal PDGFRβ expression is significantly associated with high histopathological grade, ER negativity, and shorter recurrence-free survival [66]. Indeed, some studies suggest that tumor cells secret PDGF-B as a means of recruiting/activating fibroblasts [67,68]. However, it has not yet been shown if stromal cells provide the PDGFs or what impact stromal cells may have on tumor cells with high PDGFRβ. Our study suggests breast cancer phenotypes attributed to PDGFRβ signaling can be cancer cell intrinsic. Consistent with this, some evidence suggests that breast cancer cells utilize this pathway in an autocrine fashion, producing PDGF-A/B to selfactivate PDGFRβ [40].
While studies have not examined the prognostic ramifications of tumor-cell PDGF/PDGFRβ expression, our results here using mouse and human tumors indicate a high likelihood that tumor cell PDGFRβ staining might provide prognostic information in BRCA1/INK4-RB-deficient tumors. In particular, given our results indicating that Pdgfrβ-Pkcα signaling induces molecular indicators of EMT, we anticipate that BRCA1/INK4-RB-deficient tumors might be more invasive and more likely to develop resistance to chemotherapy. Thus, we anticipate these TNBC patients may require targeted therapies individualized to the unique molecular pathways that enhance the malignancy of BRCA1/INK4-RB-deficient tumors.
Target-specific drugs are available for treating HER2-positive cancers and ER-positive luminal type cancers. Very few therapeutic options are available for highly aggressive and metastatic BLBCs. More than half of BLBCs have a dysfunctional BRCA1 pathway and harbor defects in DNA damage repair [21], which make these patients initially respond well to DNAdamaging agents such as cisplatin and PARP inhibitors. However, tumor recurrence and acquired resistance to DNA-damaging agents combine to decrease the 5-year survival of such patients [24,69]. In this report, we find preclinical evidence that Pdgfrβ-Pkcα might serve as one targetable pathway. In particular, given that pharmaceutical inhibition of PDGFRβ efficiently promotes cell death of BRCA1deficient tumor cells, it indicates the potential to tailor specific therapies to BLBC patients with BRCA1 deficiency. Thus, as a whole, this study uncovers a targetable PDGFRβ-PKCα pathway with biological and therapeutic importance to TNBC.

Conclusions
Our work offers the first genetic and biochemical evidence that PDGFRβ-PKCα signaling is repressed by BRCA1, which establishes PDGFRβ-PKCα signaling as a therapeutic target for BRCA1-deficient breast cancers. This study not only reveals the molecular mechanism of BRCA1 in suppressing EMT but also tests the efficacy of inhibitors that target PDGFRβ-PKCα signaling on suppressing BRCA1-deficient tumor initiation and progression.
Additional file 1. Primer sequence for accessing the occupancy of BRCA1 on the PDGFRβ locus.  Fig. 7b), longer exposure bands for T47D cells were shown in (B). N.S., non-specific band. (C) HCC1937 and T47D cells were treated with DMSO, PDGFR Inh III, or Ro-31-8220 at the indicated concentrations for 24 h, and the number of viable cells was determined. Data are represented as mean ± SD of triplicates. *p < 0.05 between DMSO and drug treated groups by student t test. **p < 0.01 between DMSO and drug treated groups.