Selection of gene targets and primer pairs
The set of genes represented in the RNAi library was chosen to include the vast majority of those encoding predicted signaling components and cytoskeletal regulators. Genes were selected on the basis of a combination of predictions using annotations in the FlyBase [12
] and Berkeley Drosophila
Genome Project (BDGP) databases [29
] and by BLAST searches for orthologs of known genes with functional domains via NCBI/GenBank [30
]. The selected genes were categorized according to one of the following predicted protein functions or domains: adhesion molecules, adenylate and guanylate cyclases, cytoskeletal proteins and binding proteins (such as proteins with WH and FERM domains), G proteins, GTPase-activating proteins (GAPs), GEFs, GTPases, kinases, lipid-associated proteins (such as phospholipases or proteins containing PH and PX domains), miscellaneous proteins (such as transcription factors, PI phosphotyrosine-binding domains and cell-cycle regulators), motors (such as dynein, kinesins and myosins), PDZ-domain-containing proteins, phosphatases, proteins involved in proteolysis (such as ubiquitin-conjugating enzymes and ligases), proteins containing SH2 or SH3 domains and vesicle-transport-associated proteins (such as SNAREs, SNAPs and dynamins). A complete list is presented in Additional data file 1.
Primer sequences were predicted using genomic and annotation data from the BDGP Release 1 [4
] with the Primer3 software [31
]. Primers were preferentially selected to span predicted exonic sequences if confirmed by the existence of expressed sequence tag (EST) or protein homology data. Electronic PCR [32
] was used to select amplification products from genomic sequence between 200 and 1,800 bp in length and possessing < 21 bp of exact match with any other predicted or confirmed transcript sequence. A smaller PCR product size was selected if the genomic sequence corresponded to > 500 bp coding sequence. PCR primers could only be predicted within the most proximal half of the intergenic sequence of each gene.
Generation of dsRNA
OregonR genomic DNA was PCR-amplified using Taq (PerkinElmer, Foster City, USA) with 5 μM each primer in 96-well plates (Tetrad from MJ Research Inc., Waltham, USA; 92°C for 1 min, 34 cycles of 92°C for 20 sec, 54°C for 40 sec, 72°C for 4 min, then 72°C for 3 min and held at 4°C), ethanol precipitated, washed, vacuum dried and resuspended in 7 μl DEPC-treated 100 mM Tris-HCl, 0.1 mM EDTA. Separate T3 and T7 in vitro transcription reactions were conducted (T3 and T7 MEGAscript; Ambion, Austin, USA) using 1.5 μl PCR product per well, incubated at 37°C for 4.5 h, and diluted with 47 μl of RNase-free water. T3 (50 μl) and T7 (50 μl) reaction mixes were combined, purified using RNeasy 96 Kits and a QIAvac 96 vacuum manifold (QIAGEN, Valencia, USA), soaked twice for 2 min and eluted in 80 μl RNase-free water. To anneal T3 and T7 single-strand RNAs, 50 μl purified RNA was mixed with 10 μl 6× buffer (40 mM KPO4 pH 7.5, 6 mM K-citrate pH 7.5, 4% PEG 6000) and heated in a PCR block at 68°C for 10 min and 37°C for 30 min. Purified dsRNA and remaining non-annealed mixes were stored in 96-well plates at -70°C. For screens, an average of 0.3 μg dsRNA in 3 μl was transferred from stock plates to 384-well black-sided, tissue-culture-treated optical bottom-assay plates (Corning, Acton, USA) using a multichannel pipette or a CyBio robot (CyBio US Inc., Woburn, USA).
Kc167 cells and S2R+ cells were grown in Schneider's medium (Invitrogen, Carlsbad, USA) with 10% heat-inactivated fetal bovine serum (JRH Biosciences, Fenexa, USA) and penicillin-streptomycin (Sigma, St Louis, USA) at 24°C in treated culture flasks (Falcon from BD Biosciences, Bedford, USA). S2R+ cells were removed from culture flasks using Trypsin-EDTA (Invitrogen).
RNAi and cell staining
RNAi was performed as described [6
]. Briefly, 1.2 × 104
cells in 10 μl serum-free Schneider's medium were added to dsRNAs in 384-well assay plates using a Multidrop384 liquid dispenser (Thermo Labsystems, Franklin, USA), centrifuged at 1,200 rpm for 1 min, then incubated at room temperature for 30 min before adding 30 μl more medium with serum by MultiDrop. Cells were grown for 3 days at 24°C. In the RNAi-modifier screen, 0.1 μg Pten
dsRNA in 3 μl was added to each assay-well before plating cells. Cells were processed using the MultiDrop dispenser and a multi-channel manifold (Drummond Scientific, Broomall, USA). Cells were fixed for 10 min in 4% formaldehyde (Polyscience, Niles, USA) in phosphate-buffered saline (PBS), washed twice in PBS with 0.1% Triton-X-100 (PBST), stained overnight at 4°C with FITC-conjugated anti-tubulin (DM1A; Sigma) and TRITC-phalloidin (Sigma) in PBST with 3% bovine serum albumin, stained for 10 min in PBS with 4',6-diamidino-2-phenylindole, dihydrochloride (DAPI; Sigma) and washed in PBS.
Autoscope image acquisition
Fluorescent images of cells in 384-well plates were acquired using an automated Nikon TE300 microscope with a 20× objective and HTS MetaMorph software (Universal Imaging, Downington, USA) running an automated Mac5000-driven stage, filter wheel and shutter (Ludl Electronic Products, Hawthorne, USA), an automated Pifoc focusing motor (Piezo Systems Inc., Cambridge, USA) and an Orca-ER cooled-coupled device camera (Hamamatsu Corporation USA, Bridgewater, USA). Images were also acquired using a similar automated microscope with a Prior stage and controller (instrument kindly shared by the Institute for Chemistry and Cell Biology, Harvard Medical School). Automated focusing was performed on DAPI-stained DNA. Images from UV, TRITC, and FITC channels were then collected within the same plane using preset exposures and a binning of 2 (640 w × 512 h pixels). Images from two different sites within each well were collected, representing around 12% of the total area. Multichannel images were combined as an RGB overlay within a stack of images for each plate.
Images from each channel and combined RGB images were visually scored independently by two researchers (B.B. and A.K.). Annotations assigned to each of the different sites imaged within every well were exported from MetaMorph into Excel spreadsheets. Phenotypes observed in multiple fields of replicate screens by independent observers were considered for further analysis. All visible phenotypes observed for an estimated majority of imaged cells per dsRNA treatment were recorded. Phenotypes were classified into one of seven major categories denoting visible defects in actin filaments, microtubules, DNA, cell shape, cell size, cell number and cell viability. Some descriptions were interdependent and therefore occasionally redundant: for example, cell shape was determined by an overall assessment of the actin and microtubule organization. Further subcategories were used to describe specific morphological attributes, although potentially subtle differences were still distinguishable between specific dsRNA phenotypes grouped within the same category. Specific categories included the following.
(a) Variable or undefined; (b) reduced levels or non-cortical (F-actin not apparent at the cell cortex, with diffuse or low levels of staining); (c) fibers (the appearance of spikes of F-actin away from the cortex, within the cell body); (d) puncta or dots (smaller and bigger accumulations within the cytoplasm, respectively); (e) accumulated (elevated levels and/or expanded at the cortex); (f) polarized (asymmetric distribution of actin at the cortex, usually fewer but larger accumulations than puncta or dots); (g) processes or ruffles (spiky or broad actin-rich protrusions, reminiscent of filopodia and lamellipodia).
(a) Variable or undefined; (b) reduced levels; (c) dots (as described for F-actin); (d) aberrant or frequent mitotic spindles (unusually formed or sized spindles and/or an increased frequency of spindles); (e) accumulated; (f) bipolar extensions or spikes (elongated microtubule bundles emanating as one to two opposing radial cell protrusions); (g) processes (multiple radial protrusions of microtubule bundles); (h) disorganized, uniform (a microtubule network throughout the cytoplasm, no longer with stronger staining of the perinuclear array).
(a) Variable or undefined; (b) small or condensed; (c) big or diffuse (abnormal size was estimated); (d) multinucleated cells.
(a) Variable or undefined; (b) flat; (c) retracted (pertains to S2R+ cells that remained flat but less well or less evenly spread, based on the shape and length of the cell edge and an estimate of the spreading area); (d) processes, spiky or stretchy (a description of the cell edge, in combination with F-actin and microtubule organization); (e) bipolar (pertains to Kc167 cells with a polarized axis, with varying degrees of lengthening ranging from lemon shapes to elongated spindle shapes); (f) round or nonadherent (pertains to S2R+ cells that were no longer flat).
(a) Variable; (b) small; (c) big (based on estimated size).
(a) Variable; (b) sparse (having an estimated less than half of the normal cell confluence of approximately 1,000 cells per field).
(a) Death (fewer than an estimated 100 cells per field).
Cells were plated at 106 cells per ml in 6-well plates with or without 15 μg dsRNA (results shown are either with gfp, mys, if, Rap1 or talin), as described above. After 3 days, cells in duplicate wells were processed for either protein or mRNA analyses. For protein detection on western blots, cells were washed, collected, resuspended in 75 μl lysis buffer (50 mM Tris, pH 7.5; 150 mM NaCl; 1 mM EDTA; 1% NP40; 0.5% DOC; 10% SDS; 10 mM NaF; 1 mM NaOV; protease inhibitors), incubated on ice for 15 min and spun at 4°C for 10 min before loading 10–12 μl supernatant with 2-mercaptoethanol to run on a 10% Tris-HCl polyacrylamide electrophoretic gel (BioRad, Hercules, USA). Semi-dry transfer to nitrocellulose membrane was probed with rabbit anti-Myospheroid (185E; gift of R. Hynes), mouse anti-Enabled (gift of D. Van Vactor) and mouse anti-α-tubulin (DM1A; Sigma) and detected with HRP anti-rabbit or anti-mouse (Jackson Labs, Bar Harbor, USA) with ECL Western Blotting Analysis System (Amersham Bioscience Corp., Piscataway, USA).
Alternatively, cells were lysed in 1 ml TRIzol (Invitrogen) and processed for total RNA resuspensions. Quantitated RNA samples (Bioanalyzer; Agilent Technologies, Palo Alto, USA) were normalized for reverse transcription reactions with SuperScript III (Invitrogen), then diluted cDNA was used for quantitative PCR (LightCycler FastStart DNA Master SYBR Green I, Roche Applied Science, Indianapolis, USA). Analyzed products were assayed in triplicates in multiple experiments. Individual samples were averaged, then normalized according to an adjustment factor, determined by the difference between cell types in the cross-point or cycle measurement for the rp49-positive control product. Relative levels of expression in the two cell types were presented as the difference between the averaged and adjusted cross-points (with one cycle difference approximately equivalent to a two-fold difference in expression level).