ZNF703 is a common Luminal B breast cancer oncogene that differentially regulates luminal and basal progenitors in human mammary epithelium

The telomeric amplicon at 8p12 is common in oestrogen receptor-positive (ER+) breast cancers. Array-CGH and expression analyses of 1172 primary breast tumours revealed that ZNF703 was the single gene within the minimal amplicon and was amplified predominantly in the Luminal B subtype. Amplification was shown to correlate with increased gene and protein expression and was associated with a distinct expression signature and poor clinical outcome. ZNF703 transformed NIH 3T3 fibroblasts, behaving as a classical oncogene, and regulated proliferation in human luminal breast cancer cell lines and immortalized human mammary epithelial cells. Manipulation of ZNF703 expression in the luminal MCF7 cell line modified the effects of TGFβ on proliferation. Overexpression of ZNF703 in normal human breast epithelial cells enhanced the frequency of in vitro colony-forming cells from luminal progenitors. Taken together, these data strongly point to ZNF703 as a novel oncogene in Luminal B breast cancer.


DNA and RNA array profiling in MB-series
DNA from 1001 tumors was hybridized to Affymetrix SNP 6.0 arrays to assay genomic copy number (Curtis et al., manuscript in preparation). Following correction for allelic cross-talk, probe-level normalization and summarization as implemented in the aroma.affymetrix crma function (Bengtsson et al, 2009), log2 ratios were obtained by comparison against a pooled reference. Data were then segmented using the circular binary segmentation algorithm implemented in the DNAcopy Bioconductor package (Olshen et al, 2004), with thresholds adjusted to account for differences in cellularity.
Matched mRNA was hybridized to Illumina HT-12 BeadArrays for gene-expression analysis. Hybridized BeadArrays were processed using a custom R script, which performs quality assessment and adjustment for spatial artifacts with the BASH tool (Cairns et al, 2008), followed by bead-level summarization and outlier removal as implemented in the beadarray Bioconductor package (Dunning et al, 2007). Data were normalized using a variant of quantile normalization that accounts for the reliability of probes on the HT-12 array, wherein a target distribution was derived by removing all control probes or probes that did not meet strict annotation criteria.
Samples were classified into the intrinsic subtypes based on the PAM50 gene list (Parker et al, 2009). Differential expression was performed on a subset of the normalized data that excluded any probes mapping to non-transcribed or repetitive regions. A probe-wise linear model was fitted to the data using the limma Bioconductor package (Smyth, 2004) with coefficients estimated for the contrasts of interest, with a coefficient for tissue-bank incorporated in the linear model. Here the primary contrast of interest was the comparison between Luminal B ZNF703 amplified or overexpressed versus neutral cases. After empirical Bayes' moderation of the variance, Benjamini-Hochberg adjusted p-values, log-ratios and log-odds were used to assess the evidence for differential expression.

Pathway-based analyses of MB-series tumor microarray expression data
Significantly altered pathways were identified in ZNF703-amplified versus neutral or ZNF703-upregulated versus neutral Luminal B tumors from the cohort of 1001 Page|3 primary tumors using Ingenuity Pathways Analysis (Ingenuity® Systems, ww.ingenuity.com). Genes found to be differentially expressed (p < 0.01) with a fold change value of +/-1.5 in either of these contrasts were associated with canonical pathways in Ingenuity's Knowledge Base. The significance of the association between the data set and the canonical pathway was measured using Fisher's exact test. Network building tools from the Ingenuity knowledge base were used to identify associated molecules and relationships.

Immunohistochemistry (IHC)
Antigen retrieval from formalin-fixed paraffin-embedded single tumor or tissue microarray section was achieved by heating in Citrate buffer (pH 6.0) for twenty minutes. Staining was performed on a BondTM autostainer using polyclonal rabbit anti-ZNF703 (1:25) or anti-ERLIN2 (1:1000) antibodies (Atlas Antibodies, Sweden), and bound primary antibody was detected by a polymer-conjugated secondary antibody with DAB as a chromogen. The ZNF703 antibody was validated in werstern blots of MCF-7 cells with manipulation of ZNF703 expression, a panel of cell lines with known ZNF703 copy number and NIH-3T3 cells overexpressing human ZNF703. The ERLIN2 antibody has been extensively validated by the Human Protein Atlas (www.proteinatlas.org).
Phoenix packaging cells and mouse NIH3T3 fibroblasts (ATCC) were cultured in DMEM (Invitrogen), supplemented with 10% fetal calf serum (Hyclone), 100 U/mL penicillin, and 100 µg/mL streptomycin (Invitrogen). Phoenix cells were plated 24 hr before transfection using the ProFection Mammalian Transfection System Calcium Phosphate (Promega). NIH3T3 cells were infected with retroviruses produced in the Phoenix packaging cells (24 and 48 hr after transfection) in the presence of 8 µg/ml polybrene (Sigma) and were selected with puromycin or hygromycin. Experiments were conducted within 2 weeks after infections and were performed using 2 independent NIH3T3 infected pools.

Page|4
For the proliferation assay NIH3T3 cells (2.5 × 10 4 cells) were plated in triplicate in 12-well plates and counted every day for 4 days using a Z2 Coulter (Beckman). For the focus formation assay, 10 5 cells were plated in duplicate in 6-well plates and cultured until confluence. A week later, cells were washed twice with PBS and stained with GIEMSA (Sigma). Representative pictures of the foci were taken after GIEMSA staining at low and high magnification.

MCF-7 adherent cell culture
The ER-positive breast cancer cell line MCF-7 (Soule et al, 1973)from ATCC) was cultured in DMEM supplemented with 10% v/v Fetal Bovine Serum and penicillin/streptomycin (all from Gibco, Invitrogen) according to manufacturer's recommendations. was replaced by culture medium after 6 hours, and after 48 hours protein, total RNA and chromatin were extracted from replicate samples.

Construction of lentiviral vector for ZNF703 overexpression
An 1804 bp fragment of the ZNF703 mRNA spanning the entire open reading frame and native stop codon was amplified from HCC-1500 cDNA using ZNF-for-Not1 and ZNF-rev-Not1 primers (see table below). The NotI-digested fragment was cloned into the NotI site of HIV-ZsGreen (Welm et al, 2008) and the construct verified by sequencing. The parental vector was used as a negative control. Cleared viruscontaining supernatants from transfected HEK-293T cells were obtained as described in (Welm et al, 2008) and used directly without further concentration or titration. Sub-confluent MCF-7 cultures were infected twice on consecutive days in Page|5 the presence of 1 µg/ml polybrene (Sigma). Infected cells were sorted by GFP expression using FACS Aria SORP (BD Biosciences) to eliminate non-infected cells.

Protein extraction and Western blot analysis
For protein analysis native, transfected or infected MCF-7 cultures were lysed directly in standard Laemmli Protein Sample Buffer. Samples representing equivalent cell numbers were subjected to Western Blot analysis following standard protocols using polyclonal rabbit anti-ZNF703 (Atlas Antibodies; 1:500) or AC15 monoclonal mouse anti-beta-actin (Abcam; 1:10000) antibodies and HRPconjugated rabbit-anti-goat or goat-anti-mouse (Dako) secondary antibodies, respectively, visualized by enhanced chemiluminescence using Amersham ECL Advanced Western Blotting Detection Kit (GE Healthcare).

Total RNA extraction, RT-QPCR and microarray analysis of MCF-7 cells
Total RNA was extracted using miRNeasy (Qiagen) following manufacturer's recommendations, its yield and purity monitored by spectroscopy at 260, 280 and 230 nm using the NanoDrop (Thermo Scientific) and its integrity prior to microarrays analysis verified by a Bioanalyzer trace (Agilent).
Random-primed cDNA was generated from 1 µg total RNA using either MultiScribe (Applied Biosystems) or SuperScript III (Invitrogen) according to manufacturers' recommendations. 0.5, 1 and 2 µg of a pooled sample were reverse-transcribed alongside the individual experimental samples to control for saturation of this step. A sample containing all reagents except the reverse transcriptase was included to account for genomic amplification or reagent contamination. Triplicate cDNA aliquots were subjected to real time PCR analysis (QPCR) on the ABI Prism 7900HT system (Applied Biosystems) strictly following the MIQE guidelines (Bustin et al, 2009). Each 10 ml reaction contained 1 ml cDNA, 1x SYBR Green Mastermix (Applied Biosystems) and 300 nM each forward and reverse gene-specific primers (see table below) designed to bind to different exons or over exon-exon boundaries. After an initial dissociation step of 95°C for 10 min, 40 cycles of amplification were carried out at 95°C for 15 sec and 60°C for 60 sec. The end product was assessed by a dissociation step using a ramp from 60-95°C, and all PCR products exhibited a single peak. Background readings from non reverse-transcribed RNAs and non-Page|6 template controls were at least 20-fold lower than experimental samples. Absolute expression values accounting for differences in amplification efficiency were calculated by automated software (SDS 2.3, Applied Biosystems) using linear regression of a standard titration curve included for each plate. Expression was normalized for each sample by dividing the relative expression of each gene by the geometric mean of the relative expression values of multiple internal reference genes.
Analysis of expression microarrays from lentivirus-infected and/or siRNA-transfected MCF-7 were performed in a manner similar to that described above, with the exception that data were quantile normalized. Genes with and FDR adjusted p-value < 0.05 were included in downstream geneset enrichment.

Chromatin Immunoprecipitation (ChIP)
ChIP experiments were performed as described previously (Schmidt, et al Methods, 2009) using anti-ZNF703 (sc-82451x), anti-p300 (sc-585) and anti-HDAC1 (sc-6299 sc-6298) from Santa Cruz Biotechnologies (CA, USA). Primers used for real-time PCR are listed in the table below. Statistical analysis was performed using two tailed paired T-tests. P-value cut-offs are indicated in the relevant figures.

Page|9
After 18 hours at 37°C, cells were washed three times in HBSS with 2% FBS and resuspended in 100ul of fresh Serum-Free 7. Cells were counted using a hemocytometer and equal numbers were plated in CFC assays performed as previously described (Raouf et al, 2008). For the assay, gridded 60 mm tissue culture dishes (Sarstead) were pre-coated with a 1:43 dilution of collagen solution in PBS (Stem Cell Technologies) for 1 h at 37°C and rinsed with additional PBS. Each dish was then seeded with 5000 mammary epithelial cells in 4 ml of Serum-Free 7 with 5% FBS and 1.6x10^5 freshly thawed NIH 3T3 cells, previously irradiated at 40Gy. After 8-10 days, dishes were fixed and stained with 0.8% w/vol methylene blue in methanol (Sigma). Colonies with 50 cells or more were visually scored under a dissecting microscope.

Comparison of colony formation counts of ZNF703 transfected Luminal and Basal
fractions, adjusting for empty vector control colony formation counts, was performed using ANCOVA. The data for this experiment are shown in Table S4, and ANCOVA linear model fits plotted in Figure 6B. Fold change estimates and 95% confidence intervals within each fraction, comparing ZNF703 transfected counts to control counts, were estimated using linear model fits with intercept constrained to 0.
Assumption of zero intercept was verified with ANCOVA model fit and by generating diagnostic plots. Adjustment for multiple comparisons for the fold change estimates was performed using the Bonferroni method.        B: Graphical representation of the key enzymes and molecules involved in canonical lipid metabolism and detoxification pathways enriched for amongst ZNF703-upregulated cell lines. Dark red represents molecules whose expression levels are up-regulated greater than 1.5 fold, whereas dark blue represents molecules whose expression levels are down-regulated greater than 1.5 fold. All genes/molecules are significantly differentially expressed with FDR-adjusted p-values < 0.01. Arrows represent biological relationships between molecules and the various shapes represent the functional class of the gene product. C: Barplot with enrichment for each of the network components in cell lines, where the strength of the association is represented by the -log10(p-value).

Cell-lines
Cell-lines     All Illumina probes are listed, those with an adjusted p value less that 0.05 were deemed differentially expressed. All Illumina probes are listed, those with an adjusted pvalue less that 0.05 were deemed differentially expressed. Table S3: Genes affected by ZNF703 knockdown in MCF7 cells are also differentially expressed in tumours with high vs normal ZNF703 expression. 10 out of 46 genes identified by the 50 most differentially expressed probes following ZNF703 knockdown in MCF7 cells were differentially expressed in ER positive / HER2 negative tumors with high vs normal ZNF703 expression. Tumors were divided into ZNF703 high vs normal using a threshold of 0.5 for the z-score transformed ZNF703 gene expression value (ERposHER2neg: 200 high vs 554 normal). Significance was calculated by performing a one-sided Mann-Whitney test for the expected directionality of differential expression. P-values were adjusted for multiple testing using the Bonferroni method.