Background
Triple-negative breast cancers (TNBCs) comprise mammary carcinomas that do not express estrogen receptors (ERs), progesterone receptors (PRs), and human epidermal growth factor receptor-2 (HER-2). TNBCs are an aggressive tumor subtype, characterized by a high risk of recurrence within 5 years after diagnosis and a high mortality rate [
1,
2]. Due to the lack of specific molecular targets, chemotherapy remains the standard systemic therapy for TNBCs, but it has dissatisfactory long-term results. Recently, we proposed the transmembrane protein CUB domain-containing protein-1 (CDCP1), which is overexpressed in TNBCs and involved in tumor progression, as a new therapeutic target for TNBCs [
3]. CDCP1 is a cleavable transmembrane protein that is overexpressed in several types of cancer cells [
4‐
10].
CDCP1 encodes a 135-kDa protein that is proteolyzed into a cleaved 70-kDa form [
11‐
13], which can homodimerize and initiate prometastatic activity [
11,
14].
CDCP1 increases the migration and invasiveness of cancer cells and anchorage-independent cell survival [
15,
16] through its interaction with important signalling pathways in tumor aggressiveness, such as Akt [
11], PKCδ [
17], Src [
12,
14,
16,
18‐
20], and Extracellular signal-regulated kinases 1–2 (ERK1/2) [
21]. Accordingly, several studies have suggested that the overexpression of this protein in tumors is related to worse outcomes in lung cancer [
4], pancreatic cancer [
5], renal cell carcinoma [
7], ovarian cancer [
8], and hepatocellular carcinoma [
9]. The mechanisms by which CDCP1 expression is regulated in TNBCs are unknown.
The correlation that we observed between a gain in
CDCP1 copy number and the number of cells that express CDCP1 in TNBCs supports that
CDCP1 polysomy is involved in CDCP1 overexpression in this breast cancer subtype. However, because approximately 50% of TNBC tumors overexpressing CDCP1 lack polysomy, the CDCP1 expression might be regulated by transcriptional and post-translational mechanisms, regardless of a genetic gain (e.g., by influencing the half-life of CDCP1 through EGFR-mediated inhibition of palmitoylation-dependent degradation of CDCP1 [
22]). We demonstrate that activation of platelet-derived growth factor receptors beta (PDGFRβ) by PDGF-BB upregulates CDCP1 expression and that ERK1/2 activation is crucial for this upmodulation. Consistently, a significant association between CDCP1 and PDGFRβ expression was observed in TNBC specimens, independent of gains in
CDCP1, confirming the link between these two molecules.
Methods
Cell lines, cultures, and treatments
The human breast cancer cell lines MDA-MB-231 (ATCC® HTB-26™), BT549 (ATCC® HTB122™), HCC1937 (ATCC® CRL 2336™), MDA-MB-468 (ATCC ® HTB-132™), (American Type Culture Collection, Manassas, VA), SUM149, and SUM159 (Asterand Bioscience, Detroit, MI now acquired by BioreclamationIVT, Westbury, NY) were authenticated using a panel of microsatellite markers. Cell lines were maintained at 37 °C in a humidified atmosphere of 5% CO
2 as previously described in Turdo et al. [
3]. For stimulation experiments, MDA-MB-231 cells were starved in serum-free medium for 24 h and then treated for 48 h with a pool of 5 WHFs at a final concentration of 5% as described [
23] or with PDGF-BB, Mib1b, MCP1, IP10, Il1ra, Il1b, G-CSF, Il8, Il6, EGF, FGF, Heregulin, PDGF-AA, PDGF-AB (PeproTech, Rocky Hill, NJ) at 50 ng/mL. Cells were treated in indicated experiments with c
ycloheximide (1 μM) or UO126 (2 μM), both of which were dissolved in DMSO (maximum concentration 0.1%) (Sigma-Aldrich).
Antibodies
FACS analysis was performed with Alexa Fluor® 647 anti-human CD318 (CDCP1) (BioLegend, San Diego, CA). Biochemical analyses were performed using rabbit polyclonal antibodies against CDCP1, phospho-CDCP1 (Tyr734), p44/42 MAPK (ERK1/2), phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) (Cell Signaling, Danvers, MA), and PDGFRβ (Santa Cruz Biotechnology, Dallas, TX) or mouse monoclonal anti-phospho-PDGFRβ (Tyr751) (clone 88H8) (Cell Signaling); polyclonal anti-rabbit or -mouse IgG (GE Healthcare, Chicago, IL) was the secondary antibody. Actin was revealed by probing with peroxidase-linked mouse monoclonal anti-actin (Sigma-Aldrich).
Western blot
To prepare crude cell lysates, cells were processed as described [
24]. Protein concentrations were determined by Coomassie Plus protein assay (Thermo Fisher Scientific, Waltham, MA). The samples were separated on NuPage SDS-Bis-Tris gels (Thermo Fisher Scientific) and transferred to PVDF membranes (Merck Millipore, Billerica, MA). Signals were detected using ECL reagent (GE Healthcare). Protein expression was normalized to that of actin, and densitometry was performed in Quantity One 4.6.6 (Bio-Rad, Hercules, CA).
Cytofluorimetric analysis
CDCP1 protein was detected by FACScan analysis by staining cells with PE anti-human CD318 (CDCP1) Antibody (BioLegends). Cells not stained with antibody were used as controls. The gates were set based on light scatter properties after debris and doublet exclusion; a representative gating strategy is shown in Additional file
1: Figure S1. Samples were analyzed using a FACSCalibur flow cytometer (BD Bioscineces) and FlowJo software (TreeStar).
Knockdown of PDGFRβ by siRNA transfection
To knock down PDGFRβ, cells were transfected with 100 nM of specific silencer siRNA (ID s10242) or a N.1 negative control siRNA (Thermo Fisher Scientific) using RNAiMAX (Life Technologies), harvested at 48 h post-transfection, and examined for protein expression by western blot.
Patients
Samples from 65 TNBC patients diagnosed between August 2002 and February 2007 were collected in our institute (Fondazione IRCCS Istituto Nazionale dei Tumori) [
3,
25].
Immunohistochemistry
Expression of CDCP1 and PDGFRβ was analyzed by IHC in consecutive 2-μm formalin-fixed, paraffin-embedded (FFPE) tumor sections, using rabbit polyclonal anti-CDCP1 (1:50) (PA5–17245, Thermo Fisher Scientific) and rabbit anti-human PDGFRβ (1:200) (Y92, Abcam), respectively. Antigen retrieval was performed by heating the sections for 5 min at 96 °C in 10 mM citrate buffer, pH 6.0. Staining was visualized using streptavidin-biotin-peroxidase (Dako, Agilent Technology, Santa Clara, CA) and 3,3′-diaminobenzidine (DAB; brown signal) (Dako), and the sections were counterstained with hematoxylin. Images were acquired by ECLIPSE TE2000-S inverted microscope (Nikon Instruments, Melville, NY) at 20X and 40X magnification. The reactivity of anti-CDCP1 and anti-PDGFRβ was considered to be positive per Turdo 2016 and D’Ippolito 2016 [
3,
25]. Specifically, based on the intensity of PDGFRβ staining in neoplastic cells, we assigned tumors a score of 0 (absence of signal) or 1 (weak to strong cytoplasmic signal and membrane signal). Reactivity of polyclonal anti-CDCP1 was defined as positive when ≥10% of tumor cells showed membrane staining.
Fluorescence in situ hybridization (FISH)
All FISH analyses were performed in FFPE tissues in areas that were selected by the pathologist as being CDCP1-positive by IHC or, for IHC-negative cases, representative of the tumor. Tumors were classified as positive or negative per Turdo et al. [
3].
Statistical analysis
Relationships between categorical variables were analyzed by Fisher’s exact test. Differences were considered to be significant at p ≤ 0.05. All analyses were performed using SAS 9.4 (SAS Institute Inc.).
Discussion
Our study described for the first time the role of PDGFRβ signalling in regulating CDCP1 expression in TNBCs. We have demonstrated that the induction of CDCP1 peaks on treatment with PDGF-BB, among various cytokines, chemokines, and growth factors, highlighting its significance in the regulation of this protein in the TNBC phenotype. Notably, the increase and maintenance of CDCP1 phosphorylation after treatment with PDGF-BB implicate PDGFRβ in the regulation of CDCP1 activity. Thus, we speculate that signals downstream of PDGFRβ are also crucial for the promotion of CDCP1-mediated prometastatic features through the activation of Src family kinases (SFKs). CDCP1 Tyr734 is the primary SFK-mediated phosphorylation site [
13,
29], crucial for the recruitment of PKCδ and resulting in CDCP1-mediated invasiveness.
Growth factor receptors that are involved in tumor progression have been implicated in CDCP1 overexpression. The EGF-EGFR axis promotes CDCP1 expression in ovarian cancer models, and bone morphogenetic protein 4 induces CDCP1 in pancreatic cancer cells [
5,
30]. Our analyses suggest that growth factors other than PDGF-BB mediate the upregulation of CDCP1. For example, HRG (likely via EGFR/HER3 heterodimerization) and FGF upmodulate CDCP1 on the cell membrane of TNBCs. The redundancy of signalling pathways that are downstream of these tyrosine kinase receptors suggests that the activation of several growth factor receptors converge at common mediators in CDCP1 synthesis, the most important of which are members of the RAS/RAF/MEK/ERK pathway, which have been implicated in CDCP1 mRNA and protein expression. Activation of Ras-ERK signalling alone induces CDCP1 expression in NCSLC, likely through the transcription factor AP-1 [
21].
The RAS/RAF/ERK pathway is stimulated in TNBC [
31], leading cells to acquire an aggressive phenotype—i.e., promoting invasiveness and migration [
32]. Thus, ERK1/2 might also promote these features through the regulation of CDCP1 expression. Several stimuli, mediated by growth factors, converge to activate ERK1/2, paralleling the rise in CDCP1 on the treatment of MDA-MB-231 cells with the various growth factors that we tested. As observed in a panel of TNBC models, treatment with an ERK1/2 inhibitor under standard culture conditions downregulated CDCP1, demonstrating that this pathway is crucial for CDCP1 expression in the TNBC phenotype. The MDA-MB-468 cell line was the only model that showed no variation in the CDCP1 expression on ERK1/2 inhibition; similarly, it was the TNBC line that upregulated CDCP1 RNA and protein the least on treatment with WHD [
3]. These data suggest the existence of another regulatory mechanism in addition to RTK/ERK1/2 activation. That ERK is less active when the PDGFRβ expression is abrogated in TNBC confirms the link between these two tyrosine kinases. CDCP1 expression regulation, upon PDGFRβ/ERK1/2 pathway activation, was investigated also in non-breast cancer cells. CDCP1 expression did not increase upon PDGF-BB treatment in the human large cell lung cancer cells NCI-H460 (ATCC® HTB-177™), whereas it was slightly up-regulated in the human esophageal adenocarcinoma cells OE19 (Sigma-Aldrich). Interestingly, in both cell lines, the presence of ERK1/2 inhibitor strongly reduces CDCP1 protein level (unpublished data). This data suggests that ERK1/2 could be a crucial hub for the regulation of CDCP1 expression, not only in breast cancer cells. As a support, it has been shown that CDCP1 expression is regulated through ERK1/2 recruitment in ovarian cancer cells stimulated with EGF [
29]. On the contrary, the sensitivity to PDGFRβ activation seems to be dependant by the cells.
The PDGFRβ axis is involved in breast cancer because tumor tissue and the surrounding stroma express PDGFRβ [
33,
34]. Stromal PDGFRβ expression is associated with a poor prognosis [
35,
36]. Also, breast cancer cells and fibroblasts secreted PDGF-like factors that sustain the PDGFR pathway in tumor cells [
37,
38]. Considering the low expression of PDGFRα in MDA-MB-231 cells [
39], we cannot exclude that other isoforms of PDGF receptors regulate CDCP1 expression. Notably, by immunohistochemistry, PDGFRβ and CDCP1 expression correlated significantly in a cohort of 65 TNBC specimens, confirming that the PDGF-BB/PDGFRβ axis governs CDCP1 expression in human tumors.
However, supporting that several growth factor receptors can regulate the expression of CDCP1, not all CDCP1-positive specimens expressed PDGFRβ in the tumor cells. We have reported that gains in CDCP1 are significantly associated with CDCP1 expression but that nearly half of CDCP1-positive cases do not show such gains. In the current study, in the absence of a gain in CDCP1, CDCP1 was expressed at the same frequency in PDGFRβ-positive and -negative TNBC specimens, indicating that PDGFRβ supports CDCP1 expression independently of a gain in CDCP1. Accordingly, CDCP1 polysomy was observed in MDA-MB-231 cells (data not shown), in which PDGFRβ stimulation further increased basal CDCP1 levels.
Further, our group has demonstrated that the ability of TNBCs to form vascular-like channels [
28] is associated with increased tumor aggressiveness and that this phenotype is related strictly to the expression of PDGFRβ [
25]. Considering that CDCP1 also contributes to vasculogenic mimicry [
3], we hypothesize that PDGFRβ mediated this peculiar TNBC phenotype by regulating the expression of CDCP1. Accordingly, the association between PDGFRβ and CDCP1 was nearly significant—almost 70% of vascular lacunae that expressed PDGFRβ were positive for CDCP1.
In our previous paper on CDCP1 role in TNBC [
3], we showed that knock-down of CDCP1 expression in TNBC cell lines did not affect their in vitro growth capability in 2D cultures and, accordingly, no association was found between CDCP1 expression and proliferation rates in TNBC specimens, evaluated by Ki-67 marker. Regarding the role of PDGFRβ, it is crucial for the vasculogenic properties of tumor cells, and therefore, its role in tumorigenesis mainly accounts for the activation of migration/invasion/angiogenesis pathways in cancer cells. Consistently, in a previous paper [
27], we reported that inhibition of PDGFRβ pathways only slightly influences proliferation of TNBC, and that its role in TNBC aggressiveness mainly depends on the capacity to induce vasculogenic mimicry.
In conclusion, we have identified PDGF-BB/PDGFRβ as a new pathway that is involved in the regulation of CDCP1 expression in TNCBs through ERK1/2 activation. Our results provide the basis for the potential use of PDGFRβ and ERK1/2 inhibitors in targeting the high aggressiveness of TNBCs.