Supplemental Experimental Procedures Cell culture and Sensor Ping-Pong assay. Ecotropic Phoenix HEK293T cells and Trp53 - /- MEFs (1) were grown in DMEM supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 g/ml streptomycin (100-Pen-Strep) at 37 C with 5% CO 2. ERC chicken reporter cells were grown in DMEM supplemented with 10% FBS, 1 mm sodium pyruvate and 100-Pen-Strep at 37 C with 5% CO 2, and frozen in 5% DMSO, 70% FBS and 25% culture medium. In ERCs, Tet-regulatable shrnas were induced by addition of 0.5 g/ml doxycycline (Sigma-Aldrich) that was replaced every 3 days in culture media. The Sensor Ping-Pong assay was conducted as previously described (2). All fluorescence activated cell sorting (FACS) procedures were carried out on a FACSAria II (BD Biosciences). The Top5 gating reference contained the shrnas Braf.5053, Luci.1309, Pten.1523, Pten.2049 and Trp53.1224, while the Bottom5 gating reference contained the shrnas Braf.3750, Cdkn2a.218, Cebpa.577, Cebpa.898 and Trp53.1647. ERC reporter cells were infected with the psensor shrna- Sensor libraries at single copy and sorted either after treatment with doxycycline and G418 (500 g/ml) for 6-7 days or after doxycycline and G418 withdrawal for 6-7 days. For each sort a representation of at least 1000-fold the complexity of the sorted pool was maintained. Retroviral transduction. Cells were transduced as previously described (2). Transduction efficiency was assessed 48 h post infection by quantification of fluorescent reporters using flow cytometry (Guava EasyCyte, Guava Technologies). Where a specific infection rate was desired, test infections were carried out at different dilution rates and ideal infection ratios deduced. Transduced cell populations were usually selected 48 h after infection, using 2.5 g/ml puromycin (Sigma-Aldrich) or 500 g/ml G418 (Geneticin, Gibco-Invitrogen). To obtain singlecopy integrations, cell lines (ERC, Trp53 -/- MEF) were tested to be infectable up to approximately 100% and then infected at <20% efficiency, guaranteeing <2% cells with multiple integrations (2).
Deep-sequencing based quantification of shrna representation. Assessment of library composition and changes in shrna representation over sort cycles was carried out as previously described (2). Deep-sequencing template libraries were generated by PCR amplifying shrna guide strands from vector libraries or genomic DNA isolated from at least 5 million cultured cells at each time point (>1000 fold library representation). All samples were sequenced using a custom primer reading reverse into the guide strand (mir30ecoriseq, TAGCCCCTTGAATTCCGAGGCAGTAGGCA). Individual lanes (each corresponding to 1 sample) yielded ~20-30 million reads (Illumina GA II). For each shrna and condition the number of completely matching sequences was determined, normalized to the total read number per lane and imported into a database for further analysis (Access 2007, Microsoft). As final readout of shrna potency a Sensor score was computed (Table S1). Small RNA library cloning and sequencing. Total RNA was purified, size selected, cloned and analyzed by deep-sequencing. In brief, 5-10x10 6 cells were homogenized in TRIzol (Invitrogen) followed by chloroform and phenol:chloroform:isoamyl alcohol (125:24:1, ph 4.5, Ambion) purification, precipitation in isopropanol and resuspension in RNase-free water. All samples were then DNase I (Roche) treated and again phenol:chloroform:isoamyl alcohol purified, precipitated in isopropanol and resuspended in RNase-free water. Approximately 5-10 g total RNA was run on a 12% urea polyacrylamide gel (SequaGel, National Diagnostics) to select mirna-sized small RNAs (18-28 nt) for cloning and sequencing. Mature small RNAs libraries were prepared using the TruSeq Small RNA kit (Illumina) according to manufacturer s recommendations, with index primers (Indices A, Illumina) used for pooling. PCR amplified small RNA libraries where purified on a 6% urea polyacrylamide gel (SequaGel, National Diagnostics) and analyzed by deep-sequencing (HiSeq, Illumina). Sequence reads containing the adaptor were clipped, collapsed and aligned to a reference library before normalization to the total number of reads per sample. The reference library contained guide and passenger
strands of all shrnas as well as the mouse mature mirna collection obtained from mirbase, release 19 (3). Statistical analysis. All quantifications of changes in gene expression (microarray data) were carried out using the statistical package R (version 2.15.2). Differential expression analysis for any two conditions or sample groups was conducted as follows: 1) Raw data files were imported and normalized using the RMA method of the Affy (4) package. 2) The Limma (5) package was then used to define the contrast (comparison) of interest by defining replicates as well as groups. 3) We used an empirical Bayes test to find differentially expressed genes between two groups of interest after adjusting for multiple testing using Bonferroni correction. 4) An adjusted p-value of 0.05 was used as cut-off for all tests. To evaluate the presence of potential off-target effects, two basic hypotheses were tested: 1) There exist sequence-based off-target effects caused by homology of the shrna to non-target transcripts; and 2) There exist sequence-independent off-target effects caused due to general saturation of the microrna machinery. To test for sequence-based off-target effects, we took each shrna and did a differential expression analysis (as explained above) for that shrna compared to wild-type (uninfected Trp53 -/- MEFs and Trp53 -/- MEFs infected with an empty vector control). Next, by comparing the results across the various shrnas, we identified genes that are uniquely up- or downregulated for each shrna (Figure S3A). We found differentially expressed genes for 3 shrnas at high-copy; these genes were not deregulated in the same shrna at low copy. In contrast, to test for sequence-independent off-target effects, we defined two groups for the differential expression analysis: 1) Group 1 were all the shrnas at low copy taken as replicates; and 2) Group 2 contained wild-type samples taken as replicates. The same comparison was run with all the shrnas at high copy as Group 1. While we did not observe any deregulated genes at low copy compared to wild-type, thousands of genes were deregulated at high-copy compared to wild-type (Figure 3B and S3B). For evaluation of Sensor assay performance
(Figure 1 and S1), linear dependence of variables was evaluated using the Pearson productmoment correlation coefficient r. Rank correlations were evaluated using the Spearman rank correlation coefficient (rho). Constructs and lentiviral transduction. To generate sirna-resistant KRAS expression vectors, sirna-resistant HA-tagged human KRAS WT and KRAS G12V cdna sequences were first engineered by performing QuikChange Site-Directed Mutagenesis (Agilent Technologies) and then subcloned into the tetracycline-inducible lentiviral expression vector pinducer10 (6). All the PCR-cloned sequences were verified by DNA sequencing. Viruses were produced as previously described (7) Western blot analysis. Antibodies for western blot analysis are from the following sources: human KRAS (Sigma, WH0003845M1; Santa Cruz Biotechnology, sc-30), ARAF (Santa Crus, sc-165 and sc-408), BRAF (Santa Cruz, sc-5284), RAF1 (BD Biosciences #610151 and Santa Cruz sc-7267), MEK1 (EMD Millipore, Upstate 07-641), MEK2 (Santa Cruz, sc-524), phospho-mek1&2 (Cell Signaling, #9101 and #9121), MEK1&2 (Cell Signaling, #8727), phospho-erk1&2 (Cell Signaling, #4377), ERK1&2, (Cell Signaling, #9102), p21 (BD Biosciences, 556430), BIM (Cell Signaling #2933), GAPDH (Santa Cruz, sc-25778 and sc- 47724), Beta-actin (Sigma, clone AC-74), GFP (Santa Cruz, sc-8334), and DsRed (Clontech, 632496). Rescue of KRAS knockdown. To rescue KRAS sirna-mediated cell killing, SW1116 cells stably expressing a tetracycline-inducible, sirna-resistant HA-KRAS G12V or HA-KRAS WT cdna (Table S2) were pre-treated with or without doxycycline (Sigma, 100 ng/ml) for 60h. Cells were then transfected with KRAS Sensor sirnas in the presence or absence of doxycycline. To examine MAPK pathway rescue, total cell lysates were collected 48h post-transfection and subjected to immunoblotting. For cell viability rescue, cell viability was measured 120 h posttransfection. To assess potency of KRAS shrnas, SW1116-HA-KRAS G12V cells cultured in
presence of doxycycline (2 g/ml) were infected at ~30% efficiency with one of 3 KRAS shrnas or a control shrna. Successfully transduced cells were sorted (FACS), and knockdown was quantified by immunoblotting. Nanoparticle formulation and delivery. Cyclodextrin polymer-based nanoparticle excipients are comprised of polycationic cyclodextrin polymers and adamantane-bound PEG molecules for steric stability (Calando Pharmaceuticals). Particles self-assemble when polyanionic sirnas form electrostatic interactions with the polymers at a fixed charge ratio of 3:1 (cation:anion). Some PEG molecules are conjugated to transferrin, whose receptor is overexpressed in a broad range of cancers, for receptor-mediated endocytosis by tumor cells. Additionally, due to size constraints, these intravenously administered 70 nm particles extravasate to the sites of tumor growth through leaky tumor vessels but not normal vessels. Nanoparticle-siRNA treatments were formulated by mixing equal volumes of reconsitituted CALAA-01 excipients and 4 mg/ml sirna (in 10 mm phosphate buffer + 5% (w/v) glucose). Particles were allowed to assemble for 30 min at room temperature before administration into mice via tail vein injection. FACS isolation of transduced cell populations. Tumor tissue was minced into ~1 mm pieces, resuspended in RPMI supplanted with 200 U/mL collagenase type IV and 2.5 U/mL neutral dispase and rocked at 37 C. After 30 min, 100 U/mL DNaseI was added to the solution and rocked for an additional 30 min at 37 C. Cell suspensions were passed through a 70 m filter and cells were pelleted and washed 1X with RBC lysis buffer (150 mm NH 4 Cl, 10 mm KHCO 3, 0.1 mm EDTA). Cells were pelleted and resuspended in RPMI + 1% FBS for FACS sorting. Tumor lysates. Tumor tissue was flash frozen upon harvest and briefly thawed before homogenization with a PowerGen 500 homogenizer in lysis buffer (20 mm Tris ph 7.4, 150 mm NaCl, 1% NP-40, 1 mm EDTA, 1 mm EGTA, 10% glycerol, 50 mm NaF, 5 mm NaPPi, 1 mm
PMSF, 0.5 mm DTT, 200 mm Na 3 VO 4, Sigma protease and phosphatase inhibitor solutions). Homogenized lysate was cleared by centrifugation before analysis by SDS-PAGE. Immunohistochemistry and immunofluorescence. Formalin fixed, paraffin embedded tumor tissue was cut into 5 m sections and analyzed by immunohistochemistry and immunofluorescence. Briefly, sections were de-waxed, rehydrated and blocked according to standard IHC protocols. Sections were incubated overnight using the following primary antibodies: DsRed (Clontech, 632496); GFP (Santa Cruz, sc-5386); perk1/2 (Santa Cruz, sc- 4370), ps6 (Cell Signaling, cs-4858), cleaved caspase-3 (Cell Signaling, cs-9661); Ki67 (Thermo, RM9106-S0). For immunofluorescence, sections were incubated with fluorescently conjugated secondary antibodies (Molecular Probes, A11058, A21206) and counterstained with DAPI. For immunohistochemistry, sections were incubated with biotin-conjugated secondary antibodies (Vector, BA-1000) before incubation with RTU Horseradish Peroxidase Avidin (Vector) and colorization with DAB Substrate Chromogen System (Dako). Sections were counterstained with hemotoxylin. Quantification of Ki67 and CC3-positivity was performed using GIMP software v. 2.8. ViBE analysis of phospho-proteins. Cells at ~60% confluency were transfected with 5 nm of each sirna species using forward transfections and RNAiMAX transfection reagent (Invitrogen). Transfection conditions were optimized for each cell line. Cells were lysed in lysis buffer after 72 h and lysates were cleared by centrifugation. For phosphoprotein analysis, lysates were added in triplicate to 96-well plates. Fluorescein-conjugated antibody against pmek, perk, pakt or ps6 and a matched biotin-conjugated antibody against the same phospho-protein were then added to the well. The lysate-antibody solutions were agitated on a shaking platform to allow for formation of analyte-antibody sandwiches. Detection of analyte abundance was performed with the ViBE analyzer (Bioscale, Inc.) using acoustic membrane microparticle (AMMP) technology.