Introduction
Identifying molecular targets for aggressive types of breast cancer is a milestone in the pursuit of individualized therapies. Gene-expression profiling of primary tumours has led to the following subcategories: luminal A, luminal B, the human epidermal growth factor receptor 2 (HER2) and the basal-like subtypes [
1]. Our attention was drawn to the basal-like subtype, because these tumours do not respond to available targeted therapies and patients often die within two years of diagnosis [
1,
2]. Approximately 16% of all breast cancers are basal-like [
3]; this corresponds to 46,400 women among the ~290,000 women in North America who will be diagnosed with breast cancer each year. What sets these tumours apart is that unlike many breast cancers, basal-like tumours do not express the estrogen receptor (ER) or progesterone receptor (PR), nor do they have amplified HER2. In the clinic, these tumours are often referred to as 'triple negative'. Women with triple negative tumours are not eligible for treatments that target ER (tamoxifen, aromatase inhibitors) or HER2 (trastuzumab). Instead they are treated with conventional chemotherapies, which have limited efficacy and many side effects. Therefore, it is critically important to identify alternative therapeutic strategies for basal-like breast cancer (BLBC).
We recently found that the transcription factor, Y-box binding protein-1 (YB-1), protein is commonly expressed in ER-negative breast cancers [
4], and loss of this receptor is one of the hallmarks of BLBC [
3,
5]. More recently, YB-1 was pulled out of a screen from the BLBC cell line SUM149 in an attempt to identify genes that promote malignant transformation and tumour cell growth [
6]. It has also been shown recently that epidermal growth factor receptor (EGFR) is highly expressed in approximately 50% of BLBCs [
7]. Interestingly, YB-1 was originally isolated as a transcription factor that bound to enhancer sites on the
EGFR gene, a finding that could explain, at least in part, why it promotes the growth of breast tumour cells [
8]. In keeping with this possibility, Berquin
et al. expressed YB-1 in mammary epithelial cells and observed a concomitant induction of EGFR [
6]. We demonstrated in MCF-7 (ER-positive breast cancer cells) that overexpression of YB-1 leads to an increase in the levels of EGFR mRNA and protein [
4]. This depends on the phosphorylation of YB-1 at S102 [
4]. The YB-1 S102 site is located in the DNA-binding domain, suggesting that the effect on EGFR expression was likely to be through transcriptional regulation. We demonstrated that Akt binds directly to YB-1 and phosphorylates the S102 site, an observation that was subsequently confirmed in NIH3T3 cells [
9]. We now believe that Akt is one of several kinases capable of phosphorylating the S102 site of YB-1. In support of this idea, inhibition of the kinase mTOR with rapamycin also inhibits YB-1 phosphorylation [
9]. To understand this further, we demonstrated that YB-1 binds directly to the EGFR promoter within the first 1 kb of the transcription start site, and this occurs in a phosphorylation-dependent manner [
4]. Consistent with these preclinical developments, we found that YB-1 is strongly correlated with EGFR in primary breast tumours by screening a tissue microarray of ~490 cases [
4]. More recently, we have confirmed this observation in a cohort of ~2,222 primary breast tumours. In this study, YB-1 and EGFR are once again tightly correlated (
P = 1.414 × 10
-24; data not shown).
As both YB-1 and EGFR are expressed in BLBC, we questioned whether there was a relationship between these two genes in this particular subtype of breast cancer. First, we determined whether the overexpression was caused by gene amplification, and then further dissected the regulatory relationship between the two. Finally, we addressed whether inhibiting EGFR with Iressa (also referred to as ZD1839 or gefitinib) would slow the growth of BLBC.
Materials and methods
Tumour tissue microarrays and cluster analysis
Patients in this cohort and their tumours have been previously described [
10], as have the staining conditions for YB-1, HER2, ER and PR [
10]. EGFR and CK5/6 staining was performed according to Nielsen
et al. [
7]. In total, we had interpretable data on these proteins from 285/438 total breast cancer cases. For our analysis, YB-1 scored as 0 or 1 was considered negative, and 2 or 3 was considered positive. Data was filtered to exclude patients who were missing diagnostic or survival information. Results were considered statistically significant with
P < 0.05. The data was analysed using SPSS software (Chicago, Illinois, USA).
Comparative genomic hybridization
Ten formalin-fixed and paraffin-embedded archival BLBC cases from the Vancouver General Hospital archival TMA438 series were identified based on a distinct immunohistochemical (IHC) staining pattern (ER
-, HER2
-, PR
-, CK5/6
+). Tissue cores (1.5 mm) extracted from the source blocks were treated with xylene and ethanol, as described by Garnis
et al. [
11]. Samples were placed into DNA lysis buffer comprised of 10 mM Tris, 50 mM NaCl, 1 mM EDTA, 0.5% SDS placed at 55°C, and digested with proteinase K (Invitrogen, Carlsbad, California, USA) over a period of 48 to 72 h. DNA was extracted as previously described, RNase-treated, and quantified by ND-1000 Full Spectrum UV/Vis Spectrophotometer (Nanodrop, Wilmington, Delaware, USA) [
11]. The ten BLBC specimens were assayed for genetic alterations using a whole-genome tiling path bacterial artificial chromosome (BAC) array in comparative genomic hybridization (CGH) experiments as previously described [
12]. The submegabase resolution tiling set (SMRT) array contained 32,433 overlapping BACs-derived DNA segments that provide tiling coverage over the human physical genome map. All clones were spotted in triplicate, resulting in 97,299 elements over two sides. Hybridizations were scanned using a CCD-based imaging system (Arrayworx eAuto, Applied Precision; Issaquah, Washington, USA) and analyzed using SoftWoRx Tracker Spot Analysis software as previously described [
13,
14]. Data was filtered and breakpoints were identified as previously described by Baldwin
et al. [
15]. Clones with standard deviations between replicate spots of >0.075 and with signal-to-noise ratios of <3 were filtered from raw data. Genomic breakpoint boundaries were defined by aCGH-Smooth software and visual inspection. Log 2 signal intensity ratio thresholds were used to determine regions of gain and loss, with >0.5 representing a gain and <-0.5 representing a loss.
Characterization of YB-1 and EGFR in basal-like breast cancer cells in vitro
184 htert cells were obtained from J. Carl Barrett at the US National Institute of Health, and were cultured as previously described [
16]. SUM149 cells, selected because they express markers of BLBC [
17,
18], were purchased from Astrand (Ann Arbor, Michigan, USA) and were grown according to the manufacturer's recommendation. The cells were cultured in F-12 (Ham's) media (Gibco/Invitrogen, Burlington, Ontario, USA) supplemented with 5 μg/ml insulin (Sigma, Oakville, Ontario, Canada) 1 μg/ml hydrocortisone (Sigma), 10 mM HEPES (Sigma), 5% fetal bovine serum (Gibco/Invitrogen), and 100 units/ml of penicillin/streptomycin (Gibco/Invitrogen). MDA-MB-468 cells were obtained from the ATCC and cultured in Dulbecco's modified Eagle's medium, 10% FBS and 100 units/ml penicillin/streptomycin. HCC1937 breast cancer cells, also triple negative [
19], were cultured in RPMI-1640 media supplemented with 5% FBS, 10 mM HEPES, 4.5 g/L glucose (Sigma), 1 mM sodium pyruvate (Sigma) and 100 units/ml penicillin/streptomycin. Cells were maintained at 37°C in 5% CO
2 and passaged every 2 days.
Proteins were isolated from log growing 184 htert, SUM149 and HCC1937 cells using an ELB buffer [
4]. YB-1, EGFR and actin were detected by immunoblotting. The YB-1 polyclonal antibody (courtesy of Colleen Nelson, University of British Columbia, Vancouver, Canada) was used at a dilution of 1:10,000. The EGFR monoclonal (clone 6F1, StressGen, San Diego, California, USA) and actin (Sigma, St Louis Missouri, USA) antibodies were diluted 1:1000.
Chromatin immunoprecipitation
SUM149 cells were plated at a density of 1 × 10
7 in a 150 mm dish and YB-1-promoter complexes were isolated by chromatin immunoprecipitation (ChIP) as previously described [
4]. The primers to each of the potential YB-1 binding sites were previously described [
4]. The EGFR promoter was amplified (40 cycles) using primers that span regions within the first 2 kb upstream of the start site. The input DNA was diluted fourfold before amplification.
Serial ChIP to determine YB-1 phosphorylation status
To determine whether YB-1 is serine phosphorylated at the EGFR promoter, complexes were isolated as described above with the chicken YB-1 antibody and then eluted by incubation in 10 mmol/L DTT at 37°C for 30 min with agitation. The eluate was diluted 1:50 with buffer (20 mmol/L Tris (pH 8.1), 150 mmol/L NaCl, 2 mmol/L EDTA, and 1% Triton X-100) and re-immunoprecipitated with 5 μg of anti-phosphoserine antibody (StressGen) overnight at 4°C. Secondary immunocomplexes were incubated with salmon sperm DNA/protein A agarose for 2 h at 4°C. Subsequent steps followed the ChIP protocol described previously by [
4] and PCR was performed with primers to the EGFR 2a site as described above. To test for non-specific binding species, matched IgY and IgG were incubated with an equal amount of SUM149 cross-linked DNA. The sample was then processed as described above and amplified with primers to EGFR 2a. The input DNA was also introduced as a positive control.
ChIP was also performed using a phospho-YB-1 (S102) antibody (in collaboration with Peter Mertens, Germany). The peptide sequence and supportive data demonstrating the specificity of the antibody was recently described by us [
20]. The immunoprecipitation was carried out as described above for YB-1 with protein G-agarose used in place of PreciPhen beads and rabbit IgG instead of IgY.
Electrophoretic mobility shift assay (EMSA)
Nuclear and cytoplasmic protein was extracted from log-growing SUM149 cells, MDA-MB-468 or HCC1937 cells using the NE-PER nuclear and cytoplasmic extraction reagents (Pierce Biotechnology, Rockford, Illinois, USA) following the manufacturer's protocol. Briefly, cells were centrifuged to obtain a packed cell volume and lysed in ice cold CER I with protease inhibitors. Following 5 min on ice, ice-cold CER II was added and samples centrifuged at 13,000 g for 10 min. Cytoplasmic protein was retained and the pellet re-suspended in ice-cold NER with protease inhibitors. The sample was incubated on ice for 40 min with frequent mixes and then centrifuged at 13,000 g for 10 min. The supernatant containing nuclear protein was stored. Proteins were quantified using the Bradford Assay. EMSAs were carried out using the Lightshift Chemiluminescent EMSA kit (Pierce Biotechnology), following the manufacturer's protocol. 5' Biotin-labelled complementary oligonucleotides with the following sequences, wild-type (-979 to -937) TTCACACATTGGCTTCAAAGTACCCATGGCTGGTTGCAATAAACAT, -968 mutant 5'-TTCACACCCCCGCTTCAAAGTACCCATGGCTGGTTGCAATAAACAT, -940 mutant 5'-TTCACACATTGGCTTCAAAGTACCCATGGCTGGTTGCCCCAAACAT and double mutant 5' -TTCACACCCCCGCTTCAAAGTACCCATGGCTGGTTGCCCCAAACAT were annealed to form double stranded DNA. Binding reactions consisted of 1 × binding buffer, 50 ng/μl poly dIdC, 20 fmol Biotin-labeled DNA and 5 μg nuclear protein in a 20 μl reaction. Competition reactions included 16 pmol unlabelled oligonucleotide (800-fold excess), and 1 μg chicken anti-YB-1 antibody was included to determine YB-1 involvement. An antibody to CREB (1 μg) was introduced as a negative control. The protein was incubated with the unlabelled oligonucleotide or the antibody for 20 min before the addition of the biotin-labelled oligonucleotide. The samples were incubated for 20 min at room temperature. The reaction mixture was run on a 6% non-denaturing polyacrylamide gel and transferred to a positively charged nylon membrane (Amersham Biosciences, Little Chalfont, UK). DNA was crosslinked to the membrane at 120 mJ/cm2 using a UV-light crosslinker (Stratalinker, Stratagene, La Jolla, California, USA) and detected using chemiluminescence (Pierce Biotechnology).
Tissue slices from six BLBC tumour specimens were obtained from the British Columbia Cancer Agency, Canada. Nuclear fractions were extracted using the NE-PER nuclear and cytoplasmic extraction reagents as described above. Since tissue was limited the samples were pooled before the nuclear extraction step. Electrophoretic mobility shift assays were carried out as described above with 10 μg protein.
EGFR luciferase assay
To determine whether YB-1 has a direct effect on EGFR promoter activity the normal breast cell line, 184 htert, was transfected with a 1 kb EGFR promoter construct [
21] (courtesy of Alfred C Johnson US National Cancer Institute, Massachusetts, USA), a renilla expression vector, pRL-TK (Promega, Madison, Wisconsin, USA), and a YB-1 expression plasmid, a YB-1 S102 mutant (A102) or empty vector. The cells were plated in 6-well plates (4 × 10
5 cells/well) and transfected with a total of 1.5 μg DNA using lipofectamine 2000 (Invitrogen). Cells were harvested 24 h post-transfection in 1 × PLB buffer (Promega), and luciferase activity measured. All luciferase measurements were normalized to the renilla reading from the same sample. To carry out the inverse experiment the Fast-Forward Protocol provided with the HiPerFect Transfection Reagent (Qiagen, Mississauga, Ontario, USA) was used to achieve knockdown of YB-1 in SUM149 and HCC1937 cells using small interfering RNA (siRNA) (for control and YB-1 siRNA sequences see [
4]). Briefly, cells were seeded at 4 × 10
5 cells/well of a 6-well plate in 2 ml media shortly before transfection. siRNA was diluted to 100 μl in serum-free media to achieve a final concentration of 5 nM (SUM149) or 20 nM (HCC1937), and 3 μl HiPerFect was added. Samples were vortexed, incubated at room temperature for 10 min, and then added drop-wise to the cells. At 48 h the cells were re-plated in 6-well plates (4 × 10
5 cells/well and transfected with the pER1, pRL-TK and empty vector and harvested at 24 h post-transfection as described above. Cell lysates were also collected at the time of re-plating to ensure successful knockdown of YB-1. The experiments were performed in triplicate on two separate occasions. The results are reported as the average of two experiments.
Cell viability following treatment with Iressa
SUM149 breast cancer cells were plated in 96-well plates (5 × 103 cells/well) and incubated for 24 h at 37°C in the growth media described above. Cells were treated with Iressa (isolated from tablets purchased from Astra Zeneca and kindly provided by Ching-Shih Chen, Ohio State University, USA) at the following concentrations; 0, 0.25, 0.5, 1 and 2 μM with dimethyl sulphoxide (DMSO) as vehicle control. Cell number was ascertained after 72 h treatment. Cells were washed in PBS and then incubated with Hoechst dye (1 μg/ml) for 15 mins. Nuclei counts/well were determined using the ArrayScan VTI high throughput analyser. Statistical analyses were carried out using the Student t test with significance accepted when P < 0.05.
Growth in soft agar
SUM149 cells were plated at a density of 2.5 × 10
4 in a 24-well plate in 0.6% agar, as previously described [
10] and supplemented with Iressa in the cell layer (concentrations as above). HCC1937 cells were treated with CTRL and YB-1 siRNA for 48 hours and then plated at a density of 10 × 10
3 in 0.6% agar. At the time of seeding the agar was supplemented with Iressa (0.25 to 2 μM) as described earlier. Colonies developed over 30 days and were then counted. Each experiment was performed in replicates of four and repeated twice.
EGFR sequencing from SUM149 cells
Genomic DNA was isolated from 2 × 10
7 SUM149 cells using phenol chloroform extraction followed by alcohol precipitation (modified from [
22]). DNA was quantified in a UV spectrophotometer. EGFR exons 1 to 28 were amplified by PCR and sequenced using standard techniques used by the British Columbia Cancer Agency Michael Smith Genome Sciences Centre. PCR primers were designed using human genome reference sequence acquired from the UCSC Genome Browser [
23] ([
24]) and the Primer3 program [
25]. The PCR primer sequences are listed in Additional file
1. Each PCR reaction was performed on 10 ng of SUM149 DNA and the products were visualized on a 2% agarose gel. PCR products were cleaned up using Ampure magnetic beads (Agencourt, Beverly, Massachusetts, USA) and sequenced using a standard BigDye Terminator v3.1 cycle sequencing chemistry and Applied Biosystems (Foster City, California, USA) 3730 × l DNA Analyzer.
Discussion
It has previously been reported that both YB-1 and EGFR are highly expressed in aggressive forms of breast cancer [
4,
7]. In this study we show that although these proteins are a feature of BLBC, neither gene is overexpressed owing to amplification. In further studying YB-1 as a transcription factor, we show that it transcriptionally induces EGFR in basal-like cell lines, which could lead to the increased expression observed. Importantly, we have been able to pinpoint that YB-1 binds specifically to YREs located at -968 and -940. On precisely identifying the bona fide YREs on the EGFR promoter, we demonstrate for the first time that binding to this region occurs when YB-1 is phosphorylated at S102. The high levels of both EGFR and YB-1 in BLBC begs the question of whether either of them are potential therapeutic targets. Based on the poor survival rates previously reported [
1,
2] it is clear that the BLBC subtype represents a very aggressive form of the disease, and EGFR is a rational target for the treatment of BLBC. In fact, since it was reportedly associated with this subtype of breast cancer in 2004 [
7], the use of EGFR in classifying basal-like tumours by immunohistochemistry has become widely accepted [
34,
35].
We show for the first time that the EGFR inhibitor Iressa suppresses the growth of SUM149 cells, a model for BLBC,
in vitro at concentrations achievable in patients [
28]. This is not the case for other BLBC models, as no inhibition of anchorage-independent growth was evident in the HCC1937 cells when they were treated with Iressa alone. This insensitivity is also reported in MDA-MB-468s [
30] and MDA-MB-231 cells, another triple negative cell line with high levels of EGFR expression [
36,
37]. Why the SUM149 cells alone are sensitive to the drug is not clear. Several studies suggest that activating mutations in EGFR are predictive of whether inhibitors, such as Iressa, would be effective in patients with lung cancer [
31,
38]. The same could be true for breast cancer, but it is not known whether BLBCs harbour such mutations. However, we did sequence the entire
EGFR gene from SUM149 cells and did not find activating mutations previously described for lung cancer. Whether the SNP at R521K influences sensitivity to Iressa is not known, and warrants further investigation. Another factor that may influence the sensitivity to EGFR inhibitors is the level of expression of the target itself, and also the presence of alterations in downstream signalling independent of receptor activation. For example, both the HCC1937 [
19] and MDA-MB-468 cells [
39] are PTEN null, resulting in increased propagation of the PI3-kinase pathway. She
et al. have previously shown that by inhibiting the PI3-kinase pathway with LY294002 they can sensitize cells to Iressa [
30], and we also found that by suppressing the expression of YB-1, which is downstream of phospho-Akt [
10], using siRNA in the HCC1937 cells we were able to increase the effect of Iressa. Why YB-1 sensitizes BLBC cells to Iressa is an interesting question. YB-1 has been shown to regulate the
MDR1 gene [
40,
41], and thus the P-glycoprotein pump, a member of the ABC family of transporters. This pump is involved in the efflux of many drugs, and has been associated with resistance to many chemotherapeutic agents [
42]. We recently performed a ChIP on chip analysis of YB-1 target genes in SUM149 cells, and identified ~15 ABC transporter family members that were putatively bound by YB-1, including ABCG2, ABCA5 and ABCC3. Studies carried out by Özvegy-Laczka
et al. showed that multidrug transporters such as ABCG2 may be involved in the resistance to tyrosine kinase inhibitors such as Iressa by modulating the uptake and extrusion of these drugs to and from cells [
43]. In fact, they specifically show that ABCG2, but not mutant ABCG2, protects the lung cancer cell line A431 from Iressa-induced growth inhibition [
44]. A more recent study [
45] also confirms these findings with the demonstration of decreased intracellular accumulation of low concentrations of Iressa (0.1 μM to 1 μM) and higher efflux with 1 μM Iressa. Although further work is required to ascertain the mechanism involved, the suppression of YB-1 expression could indirectly increase the levels of these inhibitors in the cells, allowing them to bind to their target and reduce cell growth.
Not withstanding that SUM149 cells are sensitive to Iressa, suggesting that some BLBCs may be also, we recognize that acquired resistance to inhibitors such as Iressa is a common problem. There are many studies that implicate the overactivation of alternative signalling pathways, such as the insulin-like growth factor 1 pathway [
46] and MET receptor amplification, leading to the activation of ERBB3–Akt pathway [
46]. Alternatively, downstream pathways can become constitutively activated, an example being KRAS, which has been reported in lung and colon cancers [
47‐
50]. Given this problem of acquired resistance, and the fact that many BLBC cases will not be sensitive, using Iressa in combination with an inhibitor for a downstream component may provide more long-term benefits.
Although we have established an association between YB-1 and EGFR in BLBC, it is likely that this transcription factor regulates the expression of other proteins linked to BLBC. For example, YB-1 regulates proliferating cell nuclear antigen (PCNA) and topoisomerase IIα [
51], both of which are expressed in BLBC [
52]. In colorectal carcinomas, YB-1 and topoisomerase IIα are co-ordinately expressed [
53]. Likewise, similar expression patterns are reported in lung cancer [
54] and synovial sarcomas [
55]. More direct evidence for this association is supported by Shibao
et al. who reported that knocking down YB-1 with antisense attenuates topoisomerase IIα reporter activity [
53]. These and other YB-1 target genes are yet to be confirmed in BLBC. If
PCNA and topoisomerase IIα are YB-1-responsive genes in BLBC, it would explain why the expression of this transcription factor is clearly associated with poor survival, based on work previously done by us [
4] and others [
56]. There are currently no commercially available inhibitors to YB-1. However, as YB-1 transactivates many growth-promoting genes, and we have shown that it can increase sensitivity to approved agents in BLBC, the question of whether it would also be a potent therapeutic target for this aggressive type of breast cancer is being actively pursued in our laboratory.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ALS carried out the luciferase assays, EMSA, phospho-YB-1 ChIP, growth assays and soft agar and was involved in drafting the manuscript. GH carried out the TMA, AA carried out the western blots on HCC1937 cells, HJ carried out the ChIP, KH was involved in acquisition of data for the growth assays, EP carried out the western blot characterising the SUM149 and HCC1937 cells, AS, TPHB and WL performed the array CGH, TON was involved in the TMA, UK and PRM made the phospho-YB-1 antibody, SA provided the primary BLBC tissue and SED conceived the studies and was involved in drafting the manuscript.