A pooled shRNA screen for regulators of primary mammary stem and progenitor cells identifies roles for Asap1 and Prox1

The preparation of mammary epithelial cell suspensions from 8–10 week-old FVB/N female mice and flow cytometric purification has been described [1]. Unless otherwise stated, all chemicals and media components were purchased from Life Technologies (Carlsbad, CA, USA) or Sigma (St Louis, MO, USA). For mammosphere culture, cells were plated in ultra-low adherence plates (Corning) in mammosphere medium (DMEM/F12 + Glutamax, 1% penicillin/streptomycin, 10 ng/ml EGF, 10 ng bFGF, 5 μg/ml insulin, 0.5 μg/ml hydrocortisone, B27 supplement) at 100,000 cells/ml and maintained as suspension cultures. Medium was exchanged every 3–4 days and spheres were passaged using trypsin-EDTA and gentle trituration every 7 days before replating at a density of not greater than 50,000 cells/ml. For irradiated NIH/3T3 (i3T3) co-culture, cells were counted manually and plated in tissue culture plates with i3T3 fibroblasts (5,000 Rads) in mammary growth medium (DMEM/F12 with glutamax, 1% penicillin/streptomycin, 10 ng/ml EGF, 5 μg/ml insulin, 0.5 μg/ml hydrocortisone, 20 ng/ml cholera toxin) with 5% FCS. After overnight incubation at 37°C in 5% O₂/5% CO₂, the cultures were changed to the same medium containing 1% FCS and incubated for a further 5 days. Colonies were harvested with trypsin/EDTA for sorting or stained with Giemsa for imaging and colony enumeration. For 3D colony assays, transduced cells were mixed with Matrigel (BD Biosciences, San Jose, CA) and cultured as described [1,11]. Cultures were imaged before fixation with 4% paraformaldehyde and embedded in paraffin. CommaDβ cells were maintained as previously described [23].

GFP-expressing transduced cells sorted by flow cytometry were manually counted and transplanted at limiting dilution as described [1], in the presence of 25% growth factor-reduced Matrigel. GFP⁺ outgrowths were visualized using a dissecting microscope (Leica Microsystems Gmbh, Wetzlar, Germany) and histology performed as described [3]. Mammary fat pad filling was quantitated comparing total fat pad area and outgrowth area using Image J software. All animal experiments conformed to regulatory standards and were approved by the Walter and Eliza Hall Institute (WEHI) Animal Ethics Committee.

A library of GIPZ plasmids expressing shRNAs was expanded individually in bacteria, then clones were pooled and plasmids purified yielding 15 pools of 90 shRNAs (Open Biosystems Transcription Factors Gene Family Library cat#RMM4950) and one pool of 63 shRNAs (Open Biosystems, custom order WEHI_73597). Lentivirus production was initiated by calcium phosphate transfection of 293T cells with pGIPZ shRNA-containing vectors and pMD2.G and psPAX2 (Addgene plasmids 12259 and 12260). Viral supernatants were collected at 26 and 44 hr post-transfection and concentrated via ultracentrifugation as per manufacturer’s protocol. Pellets were resuspended in mammary growth medium with 5% FCS, centrifuged at maximum speed for 5 minutes at room temperature to remove most serum proteins. Supernatant containing 100× concentrated virus stored at −80°C. The titre of each frozen virus stock was assessed biologically. Briefly, the day prior to transduction, 50,000 293T cells were seeded into 12-well plates in 293T medium (DMEM with 10% FCS and 1% penicillin/streptomycin). On the day of transduction, cells in three wells were counted to determine the number of cells present at transduction. Dilutions of virus made in 293T medium containing 5 μg/ml polybrene were used to replace the medium on remaining wells so that a series of wells were exposed to decreasing quantities of virus. Following transduction for 16–24 hr, the medium was changed and 48 hr later, cells were analyzed for GFP expression by flow cytometry. The number of cells at transduction and the amount of virus added to a well, where the percentage of GFP⁺ cells was between 1 and 20%, was used to calculate the transducing units (TU) per ml using the following formula: [Number of cells at transduction × (% GFP⁺ cells/100)]/volume of virus (ml). Typical TU of 1 × 10⁸/ml were achieved. For mammosphere transduction, 2 × 10⁶ purified cells were plated in mammosphere medium containing 5 μg/ml polybrene and transduced at 0 and 16 hr with 4 × 10⁶ TU. Medium was exchanged after 24 hr and a sample of the culture was taken for analysis of baseline shRNA frequency (time-point T2). Following 14 days in culture, the remaining cells were harvested (T14).

The LMS vector into which the non-silencing, and targeting shRNAs were cloned has been described [41] and the LMS-shControl, containing a non-specific shRNA sequence, was obtained from Open Biosystems. Other shRNA templates were designed as described and PCR amplified from DNA oligonucleotides using the forward primer 5’-CAGAAGGCTCGAGAAGGTATATTGCTGTTGACAGTGAGCG-3’ and the reverse primer 5’-CTAAAGTAGCCCCTTGAATTCCGAGGCAGTAGGCA-3’ [42]. Alternatively, shRNAs were purchased from OpenBiosystems as pGIPZ clones. In both cases, 137 bp shRNA products were isolated using EcoRI and XhoI digestion and subcloned into the LMS vector. Mature anti-sense sequences were as follows: shAsap1-1 (Open Biosystems clone V3LMM_492766), 5’-TTCGTCGTCATTATCTGCCTGG-3’; shAsap1-2, 5’-ATATTATATAAGTCAGCAGCCA-3’; shProx1-1 (Open biosystems clone V3LMM_506363), 5’-TTCTCTGTAACTTTCTCGG-3’; shProx1-2 (Open Biosystems clone V3LMM_506365), 5’-TTCACTCCAATGTCAACCC-3’; shSnai2-1 described previously [43]; shSnai2-2 5’-CGCAGACCCACTCTGATGTAA-3’. For production of retroviruses, Phoenix cells (10 cm plates) were transfected with 5 μg of vector using calcium phosphate precipitation and the virus supernatant was collected and filtered through a 0.45 μm filter at 48 and 72 hr. For transduction, cells on feeder layers in 6 well-plates were spin infected (2,500 rpm, 32°C, 1 hr) with 2 ml of viral supernatant per well containing 5 μg/ml polybrene. Two to four days after transduction, transduced GFP⁺ or mCherry⁺ cells were purified by flow cytometry.

Paraffin-embedded sections (5 μm) were subjected to antigen retrieval using either 10 mM citrate buffer pH6 or Tris-EDTA buffer pH9, then blocked before staining with primary antibodies overnight at 4°C and incubation with a biotinylated secondary antibody at RT for 30 min. The streptavidin-based peroxidase detection system (ABC reagent, Vector Laboratories, Burlingame, CA) was used with 3,3-diaminobenzidine as substrate (DAKO, Glostrup, Denmark). In all cases, an isotype- matched control IgG was used as a negative control.

For flow cytometry, antibodies against mouse antigens were purchased from Biolegend (San Diego, CA) or BD Biosciences (Franklin Lakes, NJ) unless otherwise specified and included CD24-PE (M1/69), CD31-B (MEC13.3), TER-119-B (TER-119), CD45-B (30-F11), CD29-FITC (HMB1-1), CD61-APC and streptavidin-APC-Cy7. For immunohistochemistry, the following antibodies were used: anti-SMA (1A4; Sigma), anti-p63 (4A4), K8/18 (Troma-1; Developmental Studies Hybridoma Bank, Iowa City), anti-Keratin 5 (Covance, Emeryville, CA), anti-Keratin 14 (LL002; Novocastra, UK), anti-E-cadherin, anti-ASAP1 (Rockland Immunochemicals Inc., Limerick, PA), anti-PROX1 (ab37128; Abcam, Cambridge, UK), anti-SNAI2 (#9585 Cell Signaling Technology, Massachussetts, USA). Secondary antibodies included biotin-conjugated goat anti-rabbit IgG, rabbit anti-rat IgG and goat anti-mouse IgG (Vector Laboratories, Burlingame, CA).

Genomic DNA was extracted using the DNeasy Blood kit (Qiagen) from samples taken at T0 and T2 timepoints and shRNA sequences were isolated from 200–800 μg gDNA (routinely 400 μg) using the PCR protocol outlined below. Primers were common to all shRNAs. The Forward primer, 5’-CAAGCAGAAGACGGCATACGAGCTCTTCCGATCTTAGTGAAGCCACAGATGTA-3’ anneals in the loop region and incorporates the P7 Illumina adapter sequence. The Reverse primer (5’-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTXXXXXGTAGCCCCTTGAATTCCGAG-3’) anneals in a region common to all shRNAs and incorporates a variable 5 bp index to enable multiplexing of samples, the P5 Illumina adapter sequence and the sequencing primer site. Following one round of PCR with an annealing temperature of 52°C and a further 30 to 32 cycles with an annealing temperature of 55°C, indexed, half shRNA products (168 bp) were pooled and sequenced on an Illumina GAIIx or HiSeq2000. Processing of the raw sequence reads was carried out in R as described previously [38]. Briefly, the number of perfect matches for each index-hairpin combination was tallied to give counts for the relevant hairpins in each sample. Downstream statistical analysis of the summarized counts was performed using the edgeR software (version 2.6.3) [44]. Outlier samples determined by visual inspection of multidimensional scaling plots were removed. An exact test for differences between the T14 and T2 biological replicate samples in each pool of shRNA was performed assuming a negative binomial distribution of the counts [45] and a common dispersion estimate. Log2-fold-changes, p-values and false discovery rates for each shRNA were reported.

Total RNA was extracted and purified from: (1) sorted luminal or basal populations from the mammary glands of female virgin 8–10 week-old FVB/N mice (three independent samples for population), (2) MaSC/basal cells cultured for 1 week under mammosphere conditions, and (3) Comma Dβ cells grown under maintenance conditions [23]. Total RNA (100 ng) was used to generate sequencing libraries for whole transcriptome analysis following the Illumina’s TruSeq RNA v2 sample preparation protocol. Completed libraries were sequenced on HiSeq 2000 with TruSeq SBS Kit v3- HS reagents (Illumina) as 100 bp single-end reads at the Australian Genome Research Facility (AGRF), Melbourne. An average of 62 million 100 bp single-end reads were obtained per sample. Reads were aligned to the mouse reference genome (mm10) using the Rsubread package (version 1.14.1) [46] and assigned to genes using the featureCounts method [47]. Data were TMM normalized [48] and transformed into log2 counts per million. Linear models with observational-level weights [49] were fitted to obtain average expression values for each gene in each sample type and moderated t-statistics were used to assess differential expression between populations [50] using the limma package (version 3.20.5) [51]. False discovery rates [52] were used together with log2-fold-changes to rank genes. These data are available through GEO Series accession number GSE63310.

For comparison of gene expression across platforms, previously published microarray data (GSE19446) [24] were compared with RNA-seq profiles. Where multiple probes were available for a given gene, the probe with the highest average expression level was taken as representative. Genes were matched between platforms using gene symbols and respective log2-fold changes were plotted.