Pancreatic adenocarcinoma generally does not respond to our current therapeutic approaches of surgical resection, radiation and chemotherapy with an overall five-year survival of only 8%. In an effort to identify potential new therapeutic approaches for pancreatic cancer, we carried out a targeted synthetic lethal CRISPR screens in the pancreatic cell line AsPC1. A custom CRISPR knockout library, termed the Therapeutic Genome (RxG) Library, which targets genes for which a current drug already exists was utilized in the screens. We used the RxG library to identify genes and corresponding gene products which modulate the sensititive and resistance to three chemotherapeutics, Gemcitabine, 5-fluorouracil, and Niraparib in the AsPC1 cell line. This screen identified multiple gene products which potentiate sensitivity to these three chemotherapeutics, and in particular, BCL2L1 which encodes BCL-xL, was identified as a common genetic factor that when mutant, enhances sensitivity to all three drugs. A novel drug, DT2216, in Phase I clinical trials which targets BCL-xL for degradation was shown to enhance the effectiveness of Gemcitabine in additional pancreatic cancer cell lines as well as patient-derived xenograft models without enhanced systemic toxicity. This work supports a possible utility of combined Gemcitabine/DT2216 as a novel therapy for pancreatic cancer.

Therapeutic Genome CRISPR (RxG) Library description 

We used online databases of existing drug targets (e. g.Drugbank, the Drug Gene Interaction Database, and the Therapeutic Targets Database) to identify candidate synthetic lethal partners for which a drug has been already identified. Each database appears to use slightly different criteria. Drugbank has 866 drug targets for which “approved” pharmaceuticals are known, while Drug Gene Interaction Database has 402 “clinically actionable” targets, and the Therapeutics Target Database identifies 397 “successful” drug targets. We combined these lists, removed duplicates and non-human targets to generate a list of 996 human target genes for which a pharmaceutical intervention currently exists. It is important to note that this set of genes does not include all potential druggable gene products (which is an evolving determination) but only a defined subset for which drugs currently exist which we term the therapeutic genome. We recognize that the criteria used by each database are different and that the specificity, strength, direction (agonist/antagonist/mixed effect) of each drug/gene product interaction are different and we refer the reader to each database for detailed description. We included the non-redundant set of drug targets from these three databases and we recognized that inclusion does not guarantee that a clinically useful intervention exists but greatly increases the likelihood that a reasonably well studied agent exists which modulates the gene product’s activity. Similarly, exclusion does not necessarily exclude the possibility that a clinically relevant drug for a particular gene product exists as new drugs are continuously being developed and the databases may have not identified an interaction. We analyzed the set of genes with the Panther Pathway Online Tool (http://pantherdb.org/) and found representation of diverse biological pathways. The top 4 ranked sgRNAs with the highest on target score and lowest off target score for each of the 996 genes were identified by the Broad Institute genetic perturbation platform (GPP) designer tool, synthesized by array synthesis (CustomArray, Inc) along with 100 non-targeting sgRNAs controls, and pooled sgRNA guides were Gibson cloned into the LentiCRISPRv2 (Zhang Lab, MIT) backbone vector. The sgRNA library was amplified and packaged into Lentivirus. We sequenced a sample of the packaged library and confirmed adequate sgRNA representation and coverage. The full list of genes and sgRNAs are presented in Supplemental excel file. 

Niraparib, 5FU, and Gemcitabine cytotoxicity

To determine the doses of Niraparib, Gemcitabine, and 5FU to use during the CRISPR screen, we assessed the cytotoxicity of these drugs in ASPC-1. Cell viability assays were carried out on 10,000 cells in 96 well plates using the CellTiter-Glo One Solution Kit (Promega) following the manufacturer’s instructions. We carried out 48 hr, 72 hr and 8 day viability at different doses of each drug. The media was replaced every 48 hrs. The cell viability was calculated as percentage of the untreated control cells and we estimated IC values for use in CRISPR screens.

Lentivirus production and titration

20 µg of the RxG library (996 genes, 4 gRNAs per gene, 100 non-targeting gRNAs) in the backbone lentiviral plasmid LentiCRISPRv2 were co-transfected in Hek293T cells with 15 µg psPAX2 (packaging plasmid) and 10 µg PMD2.G (envelop plasmid) using lipofectamine 2000 and Plus reagent in serum free Optimem medium (ThermoFisher). After 6 hours of incubation, the transfection medium was replaced with 30 ml DMEM 10% FBS, 1X Penicillin/Streptomycin, and 1% BSA (Sigma). The lentiviruses were harvested 60 Hours post transfection as follow: the medium containing the viruses was centrifuged at 3000 rpm for 10 min at 4˚C, syringe filtered through a 0.45 µm PES sterile filter (Sartorius), and the lentiviral filtrate was concentrated 40 fold using the Lenti-X Concentrator (Takara Bio) following the manufacturer’s protocol. The lentivirus pellet was resuspended in 700 µl of DMEM medium, 10% FBS and 1% BSA, aliquoted and stored at -80 C until use.

The lentivirus stock was titrated in ASPC-1 cells. In brief, 1 million of ASPC-1 cells were seeded per well (12-well microplate) 12 hours prior to the infection. The next day, the cells were transduced by spinoculation (120 min at 1000 g and 34 C) in the presence of 8 µg/ml polybrene and lentivirus stock dilutions. Cells were dissociated 24 Hrs post transduction using TryplE Express (ThermoFisher) and transferred to 6-well plates (2 wells for each dilution of lentivirus). After 24Hrs of culture, cells in one of the two wells were exposed to 2 µg/ml of puromycin HCl (ThermoFisher) for 6 days. The cell viability was determined using the MTS Kit (Promega) following the manufacturer instruction. The percentage of viability was determined using the no-puromycin corresponding wells as controls.

CRISPR screening and sequencing

To initiate the screen, 24 million ASPC1 cells were transduced as above using a MOI of 0.4 which corresponded to ~0.5 µl of concentrated lentivirus stock per million of seeded cells. Two days post spinoculation, cells were dissociated and counted. T0 samples (i.e., transduced no puromycin selection) were collected (3 replicates, 6 million cells per replicate) while the rest of cells were transferred to three T-75cm2 flasks for a period of 6 days of puromycin selection (2 µg/ml). At the end of the selection, cells were allowed to recover for one day. T₀ PURO samples (puromycin selected, time-point end of selection) were collected (3 replicates, 2 million cells per replicate) and stored at -80 C. For the drug treatment screens, twelve T-75 cm2 flasks were seeded with two million of puromycin selected cells for 12 hrs and then exposed to each drug at a sub toxic doses as follows [Niraparib (10 µM, ~IC10 72H), 5FU (2.5 µM, ~ IC5 72H,), and Gemcitabine (3.2 µM, IC30 72H)] and DMSO. Each condition (including the DMSO controls) was carried out with 3 replicates (3 flasks). The medium was changed every 3 to 4 days with fresh medium supplemented with the drugs. All cells were passaged whenever the control reached confluence, and in this case the cells were left without drugs for 24H before continuing exposure.

Amplicon preparation and Sequencing

After 16 days of exposure, two million cells for every replicate (i.e., control and treatments) were collected and the genomic DNA was extracted using the DNeasy Blood and Tissue kit following the manufacturer instruction (Qiagen). Amplicons were generated using a two steps PCR method as in Sobh et al.¹ In brief, for each sample, the gRNA region was first PCR amplified using the high fidelity Herculase II Fusion DNA Polymerase kit (Agilent) and the pair of primers CRISPR1-FOR and CRISPR1-REV (PCR1). The amplicon libraries for NGS were prepared and barcoded by carrying out a second PCR (PCR2) using a combination a common forward primer (CRISPR2-FOR) and a sample specific reverse primer (CRISPR2-REV#) incorporating respectively the P5 and the P7+barcode adapters. For each sample, two 100 µl PCR1 reactions were performed using 10 µg genomic DNA template per reaction and 18 cycles. The PCR1 products were pooled for each sample followed by two 100 µl PCR2 reactions. PCR2 were carried out using 5 µl PCR1 product per reaction, and 20 cycles producing 358 bp amplicon libraries.

The sequence of the primers used to prepare the amplicon libraries as well as the barcodes are listed in Supplemental Table 2. The amplicons were gel purified using the QIAquick Gel Extraction Kit (Qiagen) and quantified using the Qubit HS dsDNA assay (Thermoscientific). Equimolar amounts of each amplicon library were multiplexed in one pool. The library size and concentration were confirmed by TapeStation (Agilent). The sequencing was performed at the Interdisciplinary Center for Biotechnology Research (ICBR), University of Florida at Gainesville, using the NextSeq500 High Throughput single read 75 cycles platform (Illumina)

Computational Analysis

The FASTX-Toolkit was used to demultiplex raw FASTQ data which were further processed to generate reads containing only the unique 20 bp gRNA sequence. The gRNA sequences from the library were assembled into a Burrows-Wheeler index using the Bowtie build-index function and reads were aligned to the index. The efficiency of alignment was checked and the number of uniquely aligned reads for each library sequence was calculated to create a table of raw counts. Ranking of genes corresponding to perturbations that are enriched or depleted in control (DMSO) and treated cultures compared to the reference pool (T=0, post puromycin selection) was performed using a robust ranking aggregation (a-RRA) algorithm implemented in the Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout (MAGeCK) tool through the test module. Tables with raw counts corresponding to each gRNA in reference and selected samples were used as an input for MAGeCK. Gene-level ranking was based on FDRs and candidates with FDRs < 0.05 were considered as significant hits. Raw data and MAGECK analysis are provided in Supplemental excel files.

The data set contains the CRISPR screening data for AsPC1 exposed three chemotherapeutics, with a custom CRISPR Knockout library. The custom library targets 996 genes for which a currently identified drug was identified for the corresponding gene product. The complete description of this library is included in Methods and the data file, in .xls format, contains the corresponding sgRNAs for each gene.

The data file has separate tabs containing the therapeutic genome library sgRNAs, the primers used in the study, the raw counts for each sample, and the MaGeCK analysis for comparison of each treated sample with the T0 puro sample as described in methods.