The following paragraphs outline the experimental steps involved in the genome-wide RNAi screen for JAK/STAT signaling components and processing of the data to generate a short list of JAK/STAT modifiers in cultured cells:
Library generation:
Based on ~22,000 PCR fragments that cover almost the entirety of the predicted Drosophila genes (see also Hild et al., 2003), a RNAi library was generated by T7 in vitro transcription essentially as described previously (Boutros et al., 2004).
Double-stranded RNAs (dsRNA) with an average of 400bp are aliquoted in 384-well plates tissue-culture plates that are ready-to-screen. For the JAK/STAT signaling screen, we used a 'reverse bathing' protocol. RNAi in cultured Drosophila cells is performed by simply adding cells to pre-aliquoted dsRNA in serum-free medium.
Screening:
Kc167 (hemocyte-like) cells were transfected in batch with a STAT-responsive luciferase reporter (FL channel) and a constitutively-expressed Renilla co-reporter (RL channel). The co-reporter serves as a viability (not transfection) control to assess cell viability defects caused by depleting essential genes.
After 7 hours incubation at 25ºC, batch transfected cells were resuspended in serum-free medium. Subsequently 15,000 cells in 20 µl were dispensed per dsRNA containing well using an automated liquid dispenser (MultiDrop, Thermo Labsystems). Cells were incubated for 45 min and 30 µl of serum-containing medium was added to each well.
Cells were grown for 5 d to allow for protein depletion. Pathway activity was measured for using a luminescence assay for firefly and Renilla luciferase on a Mithras LB940 plate reader (Berthold Technologies).
Screens were performed in duplicate. Each plate contained dsRNA targeting stat92E, dome, hop and socs36E in A1, A2, B1, B2 which were used as positive controls (see also Supplemental Figure S1b).
Data analysis:
Each plate was normalized by median-centering FL and RL channels individually.
z-scores (number of median-adjusted standard deviations from the median of each plates) were calculated for all wells separately for FL and RL channels.
All “spiked-in” controls wells were excluded from further analysis.
Wells with significant z-scores in the FL channel were extracted if they did not show a viability phenotype (z > 2 or <-2 for FL; z < 2 and z < -2 for RL).
Hits with high variability in FL and RL channels were excluded.
Hits significant hits with viability phenotypes (Boutros et al., 2004) and phenotypes in other screened were excluded from further analysis.
Experimental validation:
Approximately 107 hits were selected for re-tests using an independent Stat92E reporter gene.
Two re-tested candidates (dBRWD3 and Ptp61F) were analyzed in vivo. Most other candidates await further characterization. A complete list with phenotypes and calculated specificities can be accessed here.
Data and Analysis files (updated):
CellScreen R package containing data files and analysis programs
[coming soon]
Sample R session:
[coming soon]