Dermcidin exerts its oncogenic effects in breast cancer via modulation of ERBB signaling

While analyzing the expression of DCD by RT-PCR in various normal and neoplastic tissues and cell lines, we identified a larger transcript co-expressed with DCD. The transcript contains a different fifth exon as a result of alternative splicing (Figure 1A), thus, we designated it DCD-SV (for DCD splice variant). This 526 bp DCD-SV encodes a 12.1 kDa protein with a different C-terminus missing the hydrophobic coiled-coil structure (amino acids 80–103) thought to be essential for the antibacterial function of DCD [2]. The expression of DCD and DCD-SV correlated well in most tissue samples and cell lines analyzed, although the relative levels of the two transcripts demonstrated some variability (Figure 1A). To define relative DCD and DCD-SV expression levels more precisely, we performed quantitative RT-PCR analysis of various human tissue samples and cell lines. Among normal tissues, placenta expressed almost only DCD-SV, whereas in normal breast both transcripts were detected at a 2:1 ratio and cell lines displayed variable DCD and DCD-SV expression levels (data not shown). Another group also identified a short truncated (DCD-SV-1) and a larger (DCD-SV-2) form of DCD in human placental tissue [19]. DCD-SV-1 is expressed in villous parenchyma whereas the larger DCD-SV-2 isoform, which is similar to the DCD-SV sequence identified in our study, is expressed preferentially in reflected membrane [16].

We performed IHC using different antibodies and routinely detected the expression of DCD and DCD-SV in epithelial cells of human eccrine sweat glands (used as control) and luminal side of secretory ducts (Figure 1B). The reactivity was not present in normal mammary epithelial cells, and reliable staining was present in membrane and weaker in cytoplasm of tumor cells (Figure 1C). Next, we examined ~600 samples of primary and invasive carcinomas spotted in two tissue microarrays slides. The patient cohort was previously clinic-pathological evaluated and the tumors classified as negative or positive for estrogen and progesterone receptors and EGFR and HER2 receptors [28]. The Nottingham system was used for assessment of histologic grade of each tumor [28]. A group of 26 samples with consistent DCD immunoreactivity in <50% or >50% of tumor cells was classified into subgroups according to their clinical and pathological features. Statistically significant associations (p < 0.05) were found between DCD reactivity >50% and the subgroups with either high histological grade or with HER2 score 3 (Table 1.A). No relationship with overall survival was found. These results are in line with the findings of our previous study analyzing a smaller cohort [1].

To further confirm the association of ERBBs and DCD expression, we compiled freely available microarray data sets of 55 human breast cancer cell lines obtained from Cancer Cell Line Encyclopedia (http://​www.​broadinstitute.​org/​ccle). The list is described in Additional file 1: Table S1 and has representative models for the different subtypes of the disease [29]. The association analyses were done across the subgroups classified as higher or lower based on whether the value was below or above the median of RMA (robust multiarray average) normalized expression value for the DCD and ERBB genes obtained in CCLE. Again, we found statistically significant association (p < 0.05) between DCD expression (RMA ≥4) with HER2 (RMA ≥8) and also with HER3 (RMA ≥9) expression (Table 1.B, Additional file 1: Table S1). As expected, in these groups are cell lines classified in the HER2 and luminal subtype, in which HER2 gene is amplified or superexpressed [29] and (Additional file 1: Table S1).

To assess the function of DCD in breast cancer cells with high endogenous expression, we generated derivatives of the MDA-MB-361 human breast cancer cell line expressing three different shRNAs against DCD (IBC-I, IBC-II, and IBC-III) using pLKO plasmid-derived lentiviruses. Efficient downregulation of DCD mRNA and protein was confirmed by multiple assays including RT-PCR analyses (2C), immuno-cytochemistry (2A), and immuno-blotting (Figure 2D). Cells expressing DCD shRNAs had significantly reduced colony-forming ability (Figure 2B).

To evaluate if down-regulation of DCD affects cellular resistance to apoptosis, we exposed the cells to various doses of cytotoxic agents and found that their cellular resistance to H₂O₂ (Figure 2E), staurosporine (Figure 2F), and TNF-α at 200 ng/ml (Figure 2G) were significantly reduced. These results are in agreement with prior studies describing higher apoptosis resistance of cancer cells overexpressing DCD [1,6,7,9].

Next, we analyzed the effect of DCD downregulation on tumorigenesis by performing xenograft assays in immunodeficient mice. Mice inoculated with MDA-MB-361 cells expressing IBC-I shRNA developed smaller tumors compared to control pLKO cells (Figure 3A). The overall weight of various organs of mice inoculated with tumor cells was significantly reduced compared to animals without tumor, but no difference in body weight was observed between control and DCD shRNA expressing cells, and we did not observe cachexia in any of the experiments (Figure 3C,E). However, we observed a significant difference in tumor mass between control pLKO and IBC-I groups, which could explain the more pronounced weight losses of carcass, gastrocnemius, and soleus skeletal muscles in these mice (Figure 3E). Macroscopic local invasion or metastasis was not observed in any of the animals analyzed (data not shown). Immunohistochemical analysis confirmed decreased DCD protein levels in DCD shRNA expressing compared to control xenografts (Figure 3D). Additionally, we observed increased expression of CK-18 in DCD shRNA expressing xenografts implying more luminal phenotype that could contribute to decreased tumor growth (Figure 3D).

To investigate the mechanisms underlying the growth and survival-promoting effects of DCD in MDA-MB-361 breast cancer cells, we analyzed the global gene expression profiles of control pLKO and DCD shRNA expressing cells (Figure 4A, Additional file 3: Table S3). Genes were identified as differentially expressed if their expression showed at least three-fold difference in each of the three pair-wise comparisons. Using these criteria we identified 208 up and 27 down-regulated genes (Additional file 3: Table S3). Down-regulation of DCD resulted in decreased levels of several genes that regulate oxidative stress, hypoxia, and angiogenesis including disulfide isomerase-associated 3, 4 and 6 (PDIA), stress 70 kDa protein chaperone (STCH), heat shock 70 kDa protein 5 (GRP78), hypoxia-inducible gene 2 protein (HIG2), VEGF-A, and VEGF-B. The c-MYC transcription factor, which controls the expression of numerous genes involved in metabolism, protein synthesis, and cell proliferation was also down-regulated in DCD shRNA expressing cells. Several genes regulating cell survival and death also showed altered expression in DCD shRNA expressing cells including protein phosphatase 3, the catalytic subunit of calcineurin A (PPP3CA), calcium/calmodulin-dependent protein kinase II delta (CAMK2D), thioredoxin-interacting protein (TXNIP), and cyclin-dependent kinase 6 (CDK-6). Calcineurin is a calcium and calmodulin-regulated protein phosphatase that acts as a molecular integrator of specific calcium signals. TXNIP is an inhibitor of thioredoxin, a central regulator of redox states [30]. Thus, cells overexpressing DCD may display increased resistance to oxidative stress-induced apoptosis due to their higher anti-oxidant activity potentially because thioredoxin is relieved of inhibition by TXNIP. Therefore, inhibiting DCD activity by antibodies or small molecules may increase tumor cell susceptibility to radiation and chemotherapy.

Systematic functional analysis of the differentially expressed genes using GEO revealed significant enrichment for genes with metabolic function among the 208 down-regulated genes, whereas among the 27 up-regulated genes were enriched in signal transduction pathways (Figure 4B). More detailed analysis of signaling networks and pathway using the Metacore software [24] predicted higher connectivity among the genes within the EGFR signaling canonical pathway. Betacellulin, amphiregulin, EGFR and c-Myc expression levels decreased in each of the three different DCD shRNA expressing cell pools compared to control pLKO (Figure 4C, Additional file 4: Excel Spreadsheet 1, Additional file 5).

To experimentally validate these predictions of network analysis studies, we analyzed the expression levels of all ERBB family members by real time PCR in control and DCD shRNA expressing cells (Figure 5A,B). These analyses indicated that the expression levels for EGFR, ERBB2, and ERBB3, and their ligands BTC, EGF, TGF-α, AREG, HB-EGF, NGR1, and NGR4 were down-regulated, whereas the ERBB4 and ligands EREG and NGR2 were up-regulated in cells expressing DCD shRNAs. The reduction of EGFR protein levels was confirmed by immunohistochemical analysis of xenografts derived from DCD shRNA expressing cells (Additional file 6). To analyze signaling changes downstream of ERBB receptors, we analyzed the phosphorylation status of EGFR, Akt and p42/44 MAPK in control and DCD shRNA expressing MDA-MB-361 cells untreated or treated with EGF 10 ng/ml. The phosphorylation of EGFR, Akt, and MAPK proteins increased in control pLKO cells but was low or no detectable in IBC-I cells (Figure 5H).

Interestingly, ERBB4 promotes differentiation in mammary epithelial cells [31,32] and it is associated with better prognosis in breast cancer patients [33,34]. ERBB4 may execute its differentiation-inducing function by dimerizing with other ERBB family members and decreasing the levels of more oncogenic ERBB heterodimers such as ERBB3/ERBB2 [35]. Correlating with this, MDA-MB-361 breast cancer cells expressing DCD shRNA displayed a more differentiated luminal epithelial cell phenotype compared to control cells (Figure 2A). The combined regulation of c-MYC, ERBBs, and several other signal pathways by DCD may play a role in this process. Thus, an intriguing hypothesis based on our data is that the physiologic function of DCD is to promote progenitor-like cellular phenotype via modulating the activity of pathways involved in maintaining stem cell states.

To further explore the relationship between DCD expression and ERBB signaling pathways, we generated derivatives of the MCF-7 estrogen receptor positive and HER2-non-amplified luminal breast cancer cells stably expressing DCD. We compared cell growth and survival in control and DCD-expressing cells as well as the expression levels of ERBB family of receptors and ligands and components of their signaling pathways (Figure 6). DCD overexpression increased colony formation and survival (Figure 6C,D) as well as xenograft growth in immunodeficient mice (data not shown). Similar observation was described previously [36]. The constitutive autocrine expression of DCD significantly increased the mRNA levels of EGFR, HER2/ErbB2, AREG, EGF, HB-EGF, NRG3, and NRG4 (Figure 6E,F). Following EGF stimulation, DCD-expressing MCF-7 cells displayed more pronounced phosphorylation of EGFR, Akt, and MAPK proteins compared to MCF-7 pCDNA cells (Figure 6G). Next, we demonstrated that the overexpression of DCD gene in SK-BR-3 cells increase the proliferation of this HER2-amplified cell line as well as tumor growth when the cells were implanted in the mammary fat pads of female immunodeficient mice (Additional file 7: Figure S3). Thus, the pattern of changes observed in these two DCD overexpressing cells are the opposite of those found in MDA-MB-361 cells expressing DCD shRNAs strengthening the link between DCD and ERBB signaling.

To demonstrate that the observed effects were due to the extracellular actions of DCD, we analyzed the proliferation of MDA-MB-361 cells treated with 1–100 nM of highly purified recombinant human DCD (rhDCD) (Figure 5C). Similar to our prior findings [1], recombinant DCD enhanced cellular proliferation at 0.1-1 nM but not at 10 and 100 nM. Similar bell-shaped dose–response curves have been observed in experiments that established Y-P30, a DCD-derived peptide, as a neural survival peptide [1,10,11], lacritin, a homolog to DCD [19] and other well-known mitogenic factors including sonic hedgehog, VEGF, FGF, and PDGF. More importantly, real time RT-PCR analyses of treated cells confirmed the upregulation of EGFR, c-MYC, EGF, HB-EGF, and NRG3 (Figure 5D,E,F), whereas the expression of HER-2, −3 and −4 receptors and ligands AREG, BTC, TGF-α, and NRG4 did not change significantly (Figure 5F,G). Finally, we tested the efficacy of trastuzumab, a humanized polyclonal antibody against HER2, and goat polyclonal antibodies against DCD for the treatment of parental MDA-MB-361 cells in vitro and in vivo (Figure 7A,B). The results demonstrated that individually or the combination of anti-DCD antibodies and trastuzumab caused a significant reduction of the number of cell colonies in cell culture as well as the tumor growth as xenograft in immunodeficient mice. These results confirm our hypothesis that DCD autocrinally produced by MDA-MB-361 cells may be acting in concert with the ligands of HER/ErbB receptor family to stimulate the growth and proliferation of breast cancer cells.