Background
Iron has been implicated in both the initiation and progression of cancer. Due to its ability to catalyze the formation of oxygen free radicals, iron can facilitate DNA damage and lead to potentially mutagenic changes in DNA [
1]. Iron can also act as a tumor growth factor, potentiating the growth of numerous tumors, including breast tumors, in animal models [
2,
3]. Consistent with these laboratory studies, epidemiologic studies have linked excess iron and cancer [
4‐
7]. For example, subjects with increased levels of circulating iron are at increased risk of cancer [
8‐
10], and conversely, subjects who have undergone phlebotomy for iron reduction are at decreased cancer risk [
7].
The major mechanism of iron import in both normal and malignant cells is the transferrin/transferrin receptor endocytic pathway. Two molecules of ferric iron bound to transferrin are endocytosed upon transferrin receptor binding. Iron is released in the acidified endosome, reduced, and imported into the cytosol, where it enters a low molecular weight, metabolically active labile iron pool (LIP). Excess iron in the cytosol is stored in ferritin or exported via the iron exporter, ferroportin [
11]. Other mechanisms of iron import include uptake of heme, ferritin, and import of siderophore-bound iron by proteins such as the secreted glycoprotein Lipocalin 2 (LCN2, NGAL), [
12‐
15].
In the duodenum, where uptake of dietary iron occurs, the mechanism of iron import involves duodenal cytochrome b (DCYTB) [
16‐
18]. Dietary iron is largely present in an oxidized form (ferric iron, Fe
+3). DCYTB acts as a ferrireductase, reducing ferric iron to ferrous iron to permit iron uptake by divalent metal transporter 1 (DMT1). Identified in 2001 [
16], DCYTB is a member of the cytochrome b561 protein family of di-heme, transplasma membrane electron transporters [
19,
20]. Reduction of iron by DCYTB is pH-dependent and ascorbate-dependent in duodenal enterocytes [
16‐
18,
21], but ascorbate-independent in bronchial epithelial cells [
22]. Copper is also a substrate for reduction by DCYTB, a reaction that occurs in a pH-independent, ascorbate-dependent manner [
18]. Additionally, DCYTB expression has been shown to maintain extracellular levels of ascorbate [
23].
Cancer cells exhibit an enhanced requirement for iron compared to their normal counterparts. To meet the increased metabolic demand for iron, breast and other cancer cells frequently increase expression of the iron importer transferrin receptor [
24‐
26]. Alternatively or additionally, cancer cells suppress expression of the iron efflux protein ferroportin [
27]. Although retained iron is sequestered in ferritin, this nevertheless results in an increase in labile iron [
27‐
29].
Measurements of the expression of genes of iron metabolism are strong predictors of patient prognosis. For example, breast cancer patient microarray data demonstrate that increased transferrin receptor expression [
30‐
32] or decreased ferroportin expression in breast tumors are associated with poor prognosis [
27]. Tumoral expression of LCN2 is also associated with poor prognosis and increased metastasis in breast cancer [
33,
34].
To ascertain which components of iron metabolism most influence breast cancer prognosis, our group studied the association of 61 “iron” genes with breast cancer patient outcome [
32]. From these analyses, an “iron gene regulatory signature” was derived, consisting of 16 genes whose expression best predicted breast cancer patient outcome. Of these 16 genes, expression of duodenal cytochrome b (DCYTB, CYBRD1, CYB561A2) was the most significantly associated with distant metastasis-free survival (DMFS), with high expression (values above the mean) associated with a reduced hazard ratio of 0.6 (
p = 1.8e-07). Since DCYTB facilitates iron import, its association with improved outcome was surprising. The expression of this gene in the breast was also unanticipated, since its best-known function involves uptake of dietary iron.
We therefore sought to understand in greater depth the nature of the association of DCYTB with breast cancer, and to explore the role of DCYTB in the breast. We first expanded our assessment of the ability of DCYTB to predict patient survival and response to therapy utilizing large, independent gene expression datasets obtained from breast cancer patients. We then investigated whether DCYTB expression influenced iron homeostasis in malignant breast cells. Our results indicate that DCYTB expression is strikingly associated with patient outcome and response to therapy. However, we found that DCYTB does not affect intracellular iron in breast cancer cells. Rather, DCYTB inhibits FAK activation and cell adhesion. These results uncouple DCYTB from iron metabolism in breast cancer tissue and provide an explanation for the paradoxical association between increased DCYTB expression and favorable prognosis in breast cancer patients.
Discussion
DCYTB was identified as one of 16 genes comprising an iron regulatory gene signature (IRGS) that is predictive of breast cancer patient survival [
32]. In the IRGS, high expression of DCYTB was associated with improved distant metastasis-free survival. This was unexpected, because in the duodenum, DCYTB acts in conjunction with DMT1 to promote iron uptake, and an extensive literature links enhanced iron uptake with increased rather than decreased cancer risk [
2‐
10]. Our results resolve this apparent paradox between the anticipated role of DCYTB and its association with favorable prognosis by revealing that in breast cancer cells, DCYTB does not play a role in iron acquisition.
We used immunohistochemical analysis to confirm the expression of DCYTB protein in breast tissue and to assess its cellular and subcellular localization (Fig.
5). We observed that DCYTB is present on the cell surface of epithelial and myoepithelial cells, and is particularly abundant at the luminal surface of ducts. DCYTB did not co-localize with DMT1, the transport protein with which DCYTB partners for uptake of iron, casting doubt on a role for DCYTB in iron transport or detoxification in breast cells (Fig.
5). We therefore used cell culture experiments to directly test the ability of DCYTB to impact iron metabolism in breast cancer cells.
Neither DCYTB overexpression nor DCYTB knockdown altered parameters of iron metabolism in breast cancer cells. Exogenously expressed DCYTB exhibited ferrireductase activity (Fig.
6b), indicating that the function of the transfected gene was preserved. However, basal levels of ferritin, an iron storage protein that is translationally regulated by iron, and transferrin receptor, an iron import protein that is posttranscriptionally regulated by iron, were unchanged following either overexpression of DCYTB in MCF7 cells (Fig.
6a) or knockdown in T47D breast cancer cells (Fig.
7a). Further, DCYTB overexpression did not affect the response of cells to excess exogenous iron (Figs.
8,
9a–c), the intracellular labile iron pool (Figs.
7b,
9b–d), or total cellular iron (Fig.
8).
To explore alternative roles for DCYTB in breast cancer, we used Signaling Pathway Impact Analysis (SPIA) as a discovery platform. We found that DCYTB exhibited a profound effect on the focal adhesion pathway, inhibiting phosphorylation of FAK, a kinase that regulates cell adhesion and motility [
54,
57] and is often aberrantly expressed in cancer [
58,
59] (Fig.
10). Phosphorylation of paxillin, an adaptor protein involved in maturation of focal adhesions, was similarly repressed by DCYTB, as was adhesion itself (Fig.
10). FAK lies at the center of a highly complex web of interacting proteins and signaling pathways [
54,
55]. A connection between DCYTB and focal adhesions has not been previously observed, and further experiments will be required to elucidate the mechanism(s) by which DCYTB influences this complex pathway.
Consistent with an inhibitory role of DCYTB on FAK activation, analysis of two combined cohorts that together total 1610 breast cancer patients as well as the GOBO cohort (n = 1379) revealed that high DCYTB expression was associated with longer distant metastasis-free survival and longer relapse-free survival (both local and distant) (Fig.
1, Additional file
1: Figure S3).
Breast cancer patients have been successfully classified into outcome groups based on molecular profiling [
37,
38], and several platforms for patient classification have been developed, including Oncotype Dx, Mammaprint, PAM50, and EndoPredict [
60,
61]. DCYTB is not included in these currently available commercial and research-based classification systems. However, we observed that DCYTB expression increased in molecular subtypes with more favorable prognosis (Fig.
3 and Additional file
1: Figure S5), demonstrating that as a prognostic marker, DCYTB exhibits behavior that mimics known molecular markers of breast cancer.
Although evaluating patient prognosis is helpful to physicians and patients, predicting outcome of therapy is equally critical to clinical decision-making, and remains a challenge in breast cancer [
60,
62,
63]. We therefore measured the association between DCYTB expression and survival in homogeneously treated groups of breast cancer patients [
40,
64]. We used two cohorts: the first was a cohort of women with ER+ tumors who had been treated with tamoxifen monotherapy (Fig.
4a), and the second was a population of women with ERRB2- tumors treated with neoadjuvant chemotherapy (Fig.
4b). We observed a significant association of DCYTB expression with DMFS and relapse. In both cohorts, patients with low DCYTB expression were more likely to recur than those with high DCYTB expression (Fig.
4a,b). These results suggest that measurement of DCYTB expression may be useful in tailoring therapy: for example, it could help guide a subset of ER+ patients to more aggressive therapy, or alternatively, identify those for whom the risks of chemotherapy are less warranted. Use of gene expression to stratify breast cancer patients in this fashion has recently shown substantial promise [
65].
Methods
Cell culture and reagents
Reagents were purchased from the following vendors: 17-β-estradiol (Sigma-Aldrich, St, Louis, MO, USA, E2758), Tamoxifen (4-hydroxy-(Z)) (EMD Millipore, Billerica, MA, USA, 579002), FerroZine (3-(2-Pyridyl)-5,6-diphenyl-1,2,4-triazine-p,p’-disulfonic acid monosodium salt hydrate) (Sigma-Aldrich, 160601), ferric ammonium citrate (Sigma-Aldrich, F5879), Doxycycline hyclate (Sigma-Aldrich, 9891), FuGENE® HD Transfection Reagent (Promega, Madison, WI, USA, E2311), hydroxyurea (Sigma-Aldrich, H8627). T47D breast cancer cells were obtained from the American Type Culture Collection (ATCC) and grown in RPMI-1640 basal medium containing 10% FBS at 37° in 5% CO
2. MCF7 breast cancer cells were obtained from the ATCC and grown in EMEM containing 10% FBS and 10 U/ml insulin. MCF10A cells were obtained from the ATCC and cultured in MEGM containing MEGM Bulletkit™ with 100 ng/ml cholera toxin (Sigma-Aldrich, C8052). SK-BR-3 were purchased from ATCC and were grown in 10% FBS in HyClone™ McCoy’s 5A Media (GE Healthcare Life Sciences, Marlborough, MA, USA). MDCK cells were a generous gift of Dr. Andrew McKie and were cultured in DMEM supplemented with 10% Tet-free FBS (Takara Bio USA, Inc., Mountain View, CA, USA, 631106 or Fisher Scientific, SH3007003T) and puromycin (1.0 ng/ml) [
18]. All basal media were obtained from Lonza (Basel, Switzerland). FBS was purchased from Gemini Bio-Products (Broderick, CA, USA).
Construction and selection of cell lines with DCYTB overexpression
Constitutive DCYTB expression vector
The DCYTB coding sequencing was amplified from cDNA of U138MG cells and cloned into BamHI and XbaI sites of the pSL2 vector, a lentiviral overexpression vector containing enhanced green fluorescent protein (EGFP) [
66]. Cloning primers were: DCYTB-F (5′ TCGGGATCCGCCATGGAGGGCTACTGGCGCT 3′) and DCYTB-R (5′ TAGTCTAGATCACATGGTAGATCTCTGCCCAG 3′). Sequence comparison with the reference gene in the NCBI database revealed that the cloned DCYTB cDNA was a polymorphic variant (S266N, rs10455 [
67]). To express wild-type DCYTB, the mutation in the pSL2-DCYTB (S266N) variant was rectified using site-directed mutagenesis. All vectors were confirmed by DNA sequencing.
Inducible DCYTB expression vector
The following primers were used to amplify human DCYTB cDNA from pSL2-DCYTB plasmid: Forward (5′-CCCTCGTAAAGAATTCGCCACCATGGCCATGGAGGGCTACTGG-3′) and reverse (5′- GAGGTGGTCTGGATCCTTACATGGTAGATCTCTGCCCAGCC-3′). Primers contained restriction enzyme sites for EcoR1 and BamH1 respectively. The PCR product of DCYTB (861 bp) was digested with EcoR1 and BamH1 and inserted between the EcoR1/BamH1 sites of the pLVX-TetOne-Puro vector (Takara Bio USA, Inc., Mountain View, CA, USA). Plasmids were purified and sequenced. Cells were transfected using FuGENE® HD transfection reagent followed by 2 weeks of puromycin selection.
siRNA
All reagents were obtained from GE Dharmacon (Lafayette, CO, USA) siDCYTB (D-17132-02 and D-17132-03) and siGAPDH (D-001140-01) were used for knockdown experiments. Transfections were performed according to the manufacturer’s recommendations using Dharmafect #1 (T-2001) transfection reagent.
Western blotting
For DCYTB analysis, non-reduced samples were used; other samples were reduced. Cells were lysed in NP-40 lysis buffer (1% Nonidet P-40, 0.5% deoxycholate, and 0.1% SDS) in the presence of protease and phosphatase inhibitors (Roche Diagnostics, Basel, Switzerland) and proteins separated by SDS-PAGE. Western blots were probed with antibodies to DCYTB (Sigma-Aldrich, HPA014757), transferrin receptor (Thermo Fisher Scientific, Waltham, MA, USA, 13-6890), ferritin H [
68], β-actin (Sigma-Aldrich, A3854), total FAK and P-FAK (Y925) (Cell Signaling Technology, Inc., Danvers, MA, USA, 13009 and 3284), phospho-paxillin (Cell Signaling Technology cat #2541), and paxillin (Cell Signaling Technology cat# 12065).
mRNA expression
qRT-PCR was performed essentially as described [
69], except that RNA was isolated and purified using the High Pure RNA Isolation Kit (Roche Diagnostics) and RT-qPCR was carried out using 2X SYBR® Green PCR Master Mix (Bio-Rad Laboratories, Inc., Hercules, CA, USA) in a ViiA7 cycler (Applied Biosystems, Inc., Foster City, CA, USA). Primers for PCR were designed with IDT PrimerQuest software (Integrated DNA Technologies, Inc., Coralville, IA, USA): DCYTB forward 5′-TGCATACAGTACATTCCCGCCAGA-3′, DCYTB reverse 5′-ATGGAACCTCTTGCTCCCTGTTCA-3′, ACTB forward 5′-TTGCCGACAGGATGCAGAAGGA-3′, ACTB reverse 5′-AGGTGGACAGCGAGGCCAGGAT-3′. GREB1 primers were as described in [
70].
Immunohistochemistry
Breast tissue microarrays were obtained from US Biomax, Inc., (Rockville, MD, USA). Antigen retrieval was performed using 0.05% citraconic anhydride (Acros Organics, Geel, Belgium) at pH 7.4 prior to immunostaining with a rabbit anti-DCYTB antibody (Sigma-Aldrich, HPA014757) or rabbit anti-DMT1 antibody (Sigma-Aldrich, HPA032140). Antibody to DCYTB was validated by immunofluorescence of cells that expressed high and low levels of DCYTB (Additional file
1: Figure S13). Slides were counterstained with hematoxylin (Poly Scientific R&D Corp., Bay Shore, NY, USA). Images were acquired using a Zeiss Axio Scan Z1 (Carl Zeiss Microscopy GmbH., Jena, Germany).To quantify DCYTB expression, stained microarray images were analyzed with Fiji software using reciprocal intensity as previously described [
71]. Briefly, diaminobenzidine (DAB) signal was isolated from images by color deconvolution. Regions of interest were drawn around epithelial tissue throughout the entire tissue core. Mean DAB intensity/area was then measured in the regions of interest (breast epithelia). Reciprocal intensity (expressed in arbitrary units) was derived by subtracting the maximum intensity value from measured mean DAB intensity/area values.
Immunofluorescence
4 × 105 DCYTB or empty vector-expressing MCF7 cells were plated in an eight-chamber slide (BD Falcon, Franklin Lakes, NJ, USA). Cells were fixed with 4% paraformaldehyde for 15 minutes at room temperature, blocked with 5% BSA at room temperature for 2 hours, and incubated with anti-DCYTB (Sigma-Aldrich cat# HPA014757) antibody overnight at 4 °C. Alexa Fluor 555 conjugated anti-rabbit IgG secondary antibody was applied at 1:800 dilutions for 1 hour. Slides were mounted with ProLong Gold anti-fade reagent (Invitrogen, Carlsbad, CA, USA). Images were acquired using inverted microscopy (Zeiss Axio Vert.A1).
Measurement of the labile iron pool (LIP)
The labile iron pool was measured essentially as described [
72]. Briefly, cells were transfected with siRNA or treated with doxycycline for 48 hours. Cells were then transferred to 96-well plates and incubated for an additional 24 hours in growth medium with or without 200 μM ferric ammonium citrate (Sigma-Aldrich, F5879) for 4 or 24 hours prior to assay. Cells were washed, incubated with 2 μM calcein acetoxymethyl ester (Life Technologies, Carlsbad, CA, USA, C1430) for 15 to 30 minutes at 37 °C, washed with phenol-free EMEM, and 100 μM starch-conjugated deferoxamine (DFO) was added (a generous gift of Biomedical Frontiers, Inc., Minneapolis, MN, USA). Fluorescence was measured at 485 nm excitation and 535 nm emission (BioTek Synergy 2, BioTek, Winooski, VT, USA). Following stabilization of the fluorescence signal, 10 μM salicylaldehyde isonicotinoyl hydrazone (SIH) was added for several minutes until a stable signal was obtained. The change in fluorescence following the addition of SIH (ΔF) was used as a measure of the labile iron pool.
Cell cycle analysis
Cells were synchronized with a 24-hour treatment of 2.0 mM hydroxyurea. Following release from synchronization, cells were removed from culture dishes and washed several times in PBS containing FBS and 2.0 mM EDTA and fixed in 70% ethanol at 4 °C overnight. Fixed cells were treated with RNase and stained with propidium iodide using FxCycle™ PI/RNase Staining Solution (Thermo Fisher Scientific, F10797). Fluorescence intensity was collected using a MACSQuant Analyzer (Miltenyi Biotec GmbH., Bergisch Gladbach, Germany). ModFit software (Verity Software House, Topsham, ME, USA) was used to calculate cell cycle histograms.
Adhesion assay
MCF7 or SKBR3 cells containing empty vector or doxycycline-inducible DCYTB were treated with 1 μg/ml doxycycline for 72 hours, trypsinized, and 20,000 cells were allowed to adhere to a 96-well plate that had been coated with fibronectin (5 μg/ml). After 1.5 hours, cells were labeled with calcein-AM (Invitrogen), non-adherent cells were washed off, and adherent cells were quantified by measuring calcein fluorescence. Each experiment was repeated three times and 8–16 replicate wells were used in each determination. Significant differences were determined using two-tailed unpaired Student’s t tests.
Microarray data sets
Cohort #1 was downloaded in October 2013 from Cancer Research [
32] as a preprocessed file. Individuals with missing data (event data was unavailable for 18 patients) were excluded from the analysis. Cohort #2 was assembled from existing databases. Criteria for Cohort #2 were a median follow-up of greater than 2.5 years, greater than 100 patients in the study, an event rate of greater than 20% and gene expression analysis on the Affymetrix (Santa Clara, CA, USA) U133 platform and an outcome measure of recurrence-free survival. Four publicly available breast cancer patient datasets met our criteria: (i) 303 (Discovery, GSE25055) and 193 (Validation, GSE25065) patients from a prospective study at M.D. Anderson Cancer Center that identified a predictive signature of response to neoadjuvant chemotherapy [
40]; (ii) a retrospective study of frozen tissue of 272 lymph node-negative patients from Rotterdam, Netherlands who did not receive systemic adjuvant or neoadjuvant therapy (GSE2034) [
73]; and (iii) 101 cancer and 14 normal patient samples from Dublin, Ireland resected prior to hormone or chemotherapy (GSE42568) [
74]. GSE25055 was downloaded April 2015 and GSE25065, GSE2034 and GSE42568 datasets were downloaded May 2015 from the National Center for Biotechnology Information Gene Expression Omnibus [
75,
76] along with clinical and follow-up data. Where possible, CEL files were downloaded, preprocessed and RMA normalized. Surrogate variable analysis (SVA package) was used to batch correct cohort #2 [
77,
78]. Analysis of the GOBO cohort was performed using online software (
http://co.bmc.lu.se/gobo). Multivariable regression analysis was performed on patients for whom all variables were included in the dataset. This restricted analysis to 612 out of 1610 patients when comparing size, grade, age and ER status, and 464 patients when the analysis included LN status. A total of 571 patients were analyzed in the GOBO cohort.
Statistical analysis
Analysis of microarray datasets was performed using R: A language and environment for computing using the affy [
79], survival [
80,
81], limma [
82] and SPIA [
51,
52] packages. Data downloaded for cohort #1 was on the Affymetrix U133A and B or U133plus2 platforms, on which two probes for DCYTB are present. In this case, the DCYTB probe with the highest absolute value of expression after normalization was used for downstream analysis. All data for cohort #2 was on the Affymetrix U133A platform, on which only one DCYTB probe is present. Kaplan-Meier (KM) survival analysis was used to determine distant metastasis-free survival (DMFS), relapse-free survival (RFS) (both local and distant) and bone-specific (RFS). Significance of KM plots was determined by the log-rank test. Cox proportional hazards regression was used to determine prognostic value of DCYTB when size, grade, age, ER status and LN status were included in the model. We used the Signaling Pathway Impact analysis (SPIA) algorithm [
51,
52], implemented in R, to identify significantly activated or inhibited pathways [pFWER (family-wise error rate) < 0.05], using information from KEGG pathway annotations and differentially expressed genes (
p < 0.05) between high and low DCYTB-expressing groups. Significance in cell culture experiments was assessed using two-tailed
t tests, with
p < 0.05 accepted as significant. Significance of DCYTB immunohistochemical staining was assessed using the Mann-Whitney rank sum test since the data were not normally distributed (Shapiro-Wilk test).
ICP-MS
All containers used for sample digestion and preparation were pretreated with trace metal grade HNO3 to remove metal contaminations. Protein samples were digested in 100 μl HNO3 (trace metal grade, Fisher Scientific) in polypropylene reagent tubes (Sarstedt, Nümbrecht, Germany) in a heating block at 90 °C for 3 hours after which 100 μl of 10 M H2O2 (trace metal grade, Fisher Scientific) was added to the solution. The digested sample was further diluted to 2 ml total volume with 1% HNO3 and stored in precleaned polypropylene tubes until measurement. To ensure elemental recovery of >90%, NIST reference material (freeze-dried, powdered bovine liver, SRM 1577c) as well as the common elemental standard mix (VHG Labs, Inc., Manchester, NH, USA) were simultaneously digested by the same method. To determine background contamination from the tubes an empty tube was treated with 1 ml HNO3 and prepared concomitantly with the samples.
Inductively coupled plasma-mass spectroscopy (ICP-MS) analysis was performed using an Agilent 7700x equipped with an ASX 250 autosampler (Agilent Technologies, Santa Clara, CA, USA). The system was operated at a radio frequency power of 1550 W, an argon plasma gas flow rate of 15 L/min, Ar carrier gas flow rate of 1.04 L/min. Elements were measured in kinetic energy discrimination (KED) mode using He gas (4.3 ml/min). Data were quantified using a 9-point (0, 0.5, 1, 2, 5, 10, 50, 100, 1000 ppb (ng/g)) calibration curve with external standards for Mg, Mn, Fe, Cu, and Zn. For each sample, data were acquired in triplicate and averaged. A coefficient of variance was determined from frequent measurements of a sample containing 10 ppb of all elements analyzed. An internal standard (Sc, Ge, Bi) introduced with the sample was used to correct for detector fluctuation and to monitor plasma stability. Elemental recovery was evaluated by measuring NIST reference material (water SRM 1643e) and found to be >90% for all determined elements.
Acknowledgements
We thank Dr. Andrew McKie for the Tet-off DCYTB-EGFP MDCK cells and for valuable discussion. We also thank Dr. Yudi Pawitan for providing additional information regarding the Stockholm cohort.