SF3B1 hotspot mutations confer sensitivity to PARP inhibition by eliciting a defective replication stress response

SF3B1 hotspot mutations are associated with a poor prognosis in several tumor types and lead to global disruption of canonical splicing. Through synthetic lethal drug screens, we identify that SF3B1 mutant (SF3B1MUT) cells are selectively sensitive to poly (ADP-ribose) polymerase inhibitors (PARPi), independent of hotspot mutation and tumor site. SF3B1MUT cells display a defective response to PARPi-induced replication stress that occurs via downregulation of the cyclin-dependent kinase 2 interacting protein (CINP), leading to increased replication fork origin firing and loss of phosphorylated CHK1 (pCHK1; S317) induction. This results in subsequent failure to resolve DNA replication intermediates and G2/M cell cycle arrest. These defects are rescued through CINP overexpression, or further targeted by a combination of ataxia-telangiectasia mutated and PARP inhibition. In vivo, PARPi produce profound antitumor effects in multiple SF3B1MUT cancer models and eliminate distant metastases. These data provide the rationale for testing the clinical efficacy of PARPi in a biomarker-driven, homologous recombination proficient, patient population.

Protein concentrations were quantified using the colorimetric Bio-Rad Protein Assay (Bio-Rad). 20 µg of cell lysate was loaded on NuPage 4-12% gradient Bis-Tris gels (Invitrogen) and proteins were separated using SDS-PAGE at 160V for 90 minutes. Proteins were transferred on to a PVDF membrane and blocked in 5% milk in TBS-T (0.1% Tween-20 in TBS) for 60 minutes. The membrane was incubated overnight in primary antibody at 4°C. The membrane was then washed three times in TBS-T before a 60-minute incubation with HRPconjugated secondary antibody. The membrane was then washed three times in TBS-T. Target protein abundance was detected using the SuperSignal™ West Pico PLUS Chemiluminescent Substrate (ThermoFisher) and an X-ray developing system. Protein expression was quantified using ImageJ. Note samples from the PiCCLe trial were exposed to 50nM talazoparib for 48 hours before western blot to mimic the conditions where CINP was identified from the Mass Spectrometry analysis of MEL202 isogenic cells.

Genome-wide CRISPR screen in K562 cells.
Doxycycline inducible K562-Cas9 cell lines were generated through the transduction of the K562 K700E cells with an Edit-R Inducible Lentiviral hEF1a-Blast-Cas9 Nuclease and subsequent selection in 10 µg/ml blasticidin. Cells were infected with a multiplicity of infection (MOI) of 0.2, with a previously validated human lentiviral gRNA CRISPR library (Addgene #67989) 36,62 . Cell lines underwent puromycin selection followed by talazoparib selection at 100nM for 24 days. DNA was extracted from surviving cells using the DNeasy Blood and Tissue Kit (Qiagen) as per manufacturers' instructions. sgRNA sequences were amplified with Q5 polymerase (NEB) with specific primers (Table S9) and sequenced on the Illumina HiSeq2500 36 . Raw read counts were converted to parts per ten million (pptm) and log2 transformed (after adding a pseudo count of 0.5) and median viability effect z-scores (the rate of decrease in abundance of each sgRNA in the population over time in the absence of drug treatment) calculated for all guides covering a gene and ranked according to z-score 36 .

Splice variant analysis by qPCR.
The analysis of alternatively spliced exons was performed using 384-well plates using SYBR Green reagents (Invitrogen) on an ABI 9700HT. Primers were designed (http://primer3.ut.ee) for sequences of predicted, conical and alternative spliced exons 3 . Relative abundance of spliced and un-spliced transcript was defined by 2 −(CT variant1 mRNA -CT variant2 mRNA) and statistical significance was evaluated through a two-tailed Mann-Whitney U-test for each splicing event.
Primers are listed in Table S8.

Ex vivo talazoparib efficacy studies.
The efficacy of talazoparib treatment on organoid models (ex vivo, 3D Matrigel assay) for the selected PDO models, SUM149 cell lines and the subsequent PDX11310 treatment in vivo study was carried out by Crown Bioscience San Diego. Briefly, 10 NOD-SCID mice were inoculated with each of the PDX models and tumours between 500-800 mm 3 were harvested and transferred on ice immediately. Tumours were processed using MACS tumour dissociation kit (Cat no 130-095-929) according to the manufacturer's protocol. Cells were seeded in 3D Matrigel assays in which cells were treated with the compounds at 9 doses in triplicate. Each model was tested with test articles and a reference drug control (cisplatin). Viability was measured using Cell Titre Glo after 7 days of drug exposure.   Figure 2. SF3B1 hotspot mutations confer transcriptional alterations. d, Boxplots of percent spliced in (PSI) ratio in K562 RNA-sequencing data of the specific alternatively spliced junction of transcripts that are consistently aberrantly spliced in multiple patient data sets between SF3B1 WT and SF3B1 MUT primary tumours 1 and the log 2 protein abundance identified from the massspectrometry data from MEL202 isogenic cell line model. i) BRD9, ii) SKIV2L, iii) DPH5, iv) EI24 and v) OXA1L. data are from n=3 independent biological repeats. e, Representative micrographs of CINP immunohistochemistry of MEL202 R625G-DEG siCINP, MEL202 R625G-DEG and MEL202 R625G paraffin embedded cell pellets (scale bar = 200µm). Images are representative of at least n=4 independent biological replicate experiments. f, Barplot showing CINP RNA levels in the MP41 SF3B1 WT UM cells and MEL202, K562 and NALM-6 series of isogenic cell lines, normalised to b-Actin. Pair-wise t-tests with Tukey's multiple correction. Data are mean +/-s.e.m. of n=4 technical replicates.