Page 87 - 2022_01-Haematologica-web

Mechanisms of sorafenib resistance in AML
(Biorad), transferred onto polyvinylidene fluoride membranes (Amersham), and subjected to immunoblotting using primary antibodies from Cell Signaling Technologies: p44/42 MAPK (ERK1/2; #9102), phospho-ERK1/2 (phospho-p44/42 MAPK Thr202/Tyr204; #4376), AKT (#9272), phospho-AKT (Ser473; #4060), TSC1 (#6935), phospho-TSC2 (Ser664; #40729), phospho- TSC2 (Tyr1571; #3614), TSC2 (#4308), phospho-mTOR (Ser2481; #2974), phospho-mTOR (Ser2448; #2971), mTOR (#2983), NF1 (#14623), MEK (#9122S), phosphor-MEK (#9154), vinculin (#4650); from ThermoFisher Scientific: GAPDH (#AM4300); from Millipore: anti-pan_Ras (clone RAS 10 MABS195) from Sigma: LZTR1 (HPA071248). Corresponding horseradish perodixase-con- jugated secondary antibodies (Promega) were used for chemilumi- nescent detection.
Biostatistical analysis
The bioinformatics pipeline for analyzing CRISPR library sequences was MAGeCK (model-based analysis of genome-wide CRISPR-Cas9 knockout).34 The hits were prioritized according to a previously described tiering structure.32 Briefly, tier 1 represents hits having a log2 fold change ≥2, 75% of sgRNA per gene present and concordance among sgRNA per gene ≥75%; tier 2 hits have a log2 fold change ≥2 and concordance among sgRNA per gene of 100%; tier 3 hits have a log2 fold change ≥1 and a concordance among sgRNA per gene of 100%. Singleton hits represent signifi- cantly enriched genes with log2 fold change ≥2, an adjusted sgRNA count of 1 and average control mean ≥100 reads. Enriched hits not satisfying these criteria were classified into the unassigned group.
Data availability.
Raw data files for CRISPR screens have been deposited at GEO and can be found under the accession number GSE138343.
Results
The MTOR and MAPK pathways are central components in resistance to sorafenib
To identify genes whose loss-of-function variants con- tribute to resistance to sorafenib in AML, we selected MOLM13 cells, an AML cell line harboring an FLT3-ITD mutation resulting in sensitivity to several FLT3 inhibitors, including sorafenib. MOLM13 cells, engi- neered to express Cas9, were stably transduced with a genome-wide lentiviral sgRNA CRISPR knockout library33 and treated for 14 days with vehicle or 50 nM sorafenib, a concentration projected to kill 80% of the cells within 3 days of drug administration (IC80). Genomic DNA was harvested from control and sorafenib-treated cultures and evaluated for enriched sgRNA using MAGeCK robust rank aggregation (RRA) analyses34 (Figure 1A, B; Online Supplementary Table S1).
A comparison of sequencing reads from sorafenib- treated cultures and vehicle-treated controls identified significant enrichment for sgRNA targeting negative reg- ulators of the MAPK and AKT/MTOR pathways (Figure 1B-D). The screen uncovered negative RAS/RAF/MEK/ERK regulator, leucine zipper like tran- scription regulator 1 (LZTR1), which inhibits the MAPK pathway by regulating RAS ubiquitination and degrada- tion.35,36 a negative regulator of RAS signaling, neurofi- bromin 1 (NF1); three components of the tuberous scle- rosis (TSC) complex including TSC complex subunit 1 (TSC1), TSC complex subunit 2 (TSC2), and TBC1 domain family member 7 (TBC1D7).37 Top hits also
included members of the GATOR1 complex, encoded by NPRL2 and DEPCD5.38 To prioritize candidates for vali- dation, we developed a tiering structure that incorpo- rates three key factors: evidence (determined by the num- ber of sgRNA guide hits per gene), concordance (indicated by the agreement across the set of guides for a given gene) and discovery (based on effect size) to rank sgRNA hits and enable a progression to pathway analysis for lower scoring hits.38 Using the prioritization scheme, the tier 1 hits (n=16) included LZTR1, TSC2, and TBC1D7 and several genes implicated in RNA splicing and ribo- some biogenesis, such as DHX15, EBNA1BP2, LSM5, PUS7, RPSA and ABCB1 transporter, linked to poor prog- nostic factors in AML (Online Supplementary Tables S1 and S2, Online Supplementary Figure S1). Our tier struc- ture imposed additional constraints for ranking sgRNA hits into the more selective tiers, which generally pre- served MAGeCK RRA rankings, although there were exceptions such as TSC1, which ranked as a tier 3 hit because of the variance in its log-fold change across the set of sgRNA for this gene. Using a false discovery rate cutoff, we decided to focus here on connecting the AKT/PI3K/MTOR and RAS/MAPK/MEK networks to verify candidates emerging from the screen (Figure 1B bottom panel, D).
Deficiency of top hit genes decreases sensitivity to sorafenib in acute myeloid leukemia cell lines
To validate top hit genes belonging to the LZTR1-con- nected network, we transduced MOLM13 cells with lentivirus expressing Cas9 and individual sgRNA to gener- ate cells deficient in single genes. Sensitivity to sorafenib was assessed in 72 h cell viability MTS assays. Cells in which LZTR1, NF1, TSC1, TSC2, or NPRL2 were inacti- vated showed reduced sensitivity to sorafenib (Figure 1E). The degree of resistance to sorafenib varied across target- ed genes, with TSC1- and LZTR1-deficient cells demon- strating the strongest resistance to sorafenib (parental IC50 = 5.03 nM, NT (non-targeting control) IC50 = 6.31 nM, sgTSC1 IC50 = 97.34 nM, and sgLZTR1 IC50 = 22.37 nM), while targeting of TSC2 yielded comparably more modest resistance to sorafenib (IC50 = 14 nM) (Figure 1E). TBC1D7- deficient cells had decreased sensitivity to sorafenib while DEPCD5-deficient cells were modestly resistant (Online Supplementary Figure S2). Deficiencies of LZRT1, NF1, TSC1, TSC2 and NPRL2 were evident by western blot analysis (Online Supplementary Figure S3A). The correspon- ding efficiencies of CRISPR knockouts were determined using Inference of CRISPR Edits (ICE) software (Synthego.com) (Online Supplementary Figure S3B).
Reduced expression levels of LZTR1, NF1, TSC1, and TSC2 correlate with reduced sensitivity to sorafenib in samples from patients with acute myeloid leukemia and deficiency results in hyperactivation of MAPK or MTOR pathways in acute myeloid leukemia cells
We evaluated results from our CRISPR screen for rele- vance to drug sensitivity and gene expression profiles observed in patients’ samples in the Beat AML database.3 RNA expression levels of LZTR1, NF1, and TSC2 showed negative correlations with sensitivity to sorafenib in sam- ples from AML patients harboring FLT3-ITD mutations (P<0.0001, P<0.001, and P<0.01, respectively) (Figure 2A). We did not observe a significant negative correlation between gene expression and sensitivity to sorafenib for
haematologica | 2022; 107(1)
79