Reversal of malignant ADAR1 splice isoform switching with Rebecsinib

Abstract

Adenosine deaminase acting on RNA1 (ADAR1) preserves genomic integrity by preventing retroviral integration and retrotransposition during stress responses. However, inflammatory-microenvironment-induced ADAR1p110 to p150 splice isoform switching drives cancer stem cell (CSC) generation and therapeutic resistance in 20 malignancies. Previously, predicting and preventing ADAR1p150-mediated malignant RNA editing represented a significant challenge. Thus, we developed lentiviral ADAR1 and splicing reporters for non-invasive detection of splicing-mediated ADAR1 adenosine-to-inosine (A-to-I) RNA editing activation; a quantitative ADAR1p150 intracellular flow cytometric assay; a selective small-molecule inhibitor of splicing-mediated ADAR1 activation, Rebecsinib, which inhibits leukemia stem cell (LSC) self-renewal and prolongs humanized LSC mouse model survival at doses that spare normal hematopoietic stem and progenitor cells (HSPCs); and pre-IND studies showing favorable Rebecsinib toxicokinetic and pharmacodynamic (TK/PD) properties. Together, these results lay the foundation for developing Rebecsinib as a clinical ADAR1p150 antagonist aimed at obviating malignant microenvironment-driven LSC generation.

Keywords: ADAR1; RNA editing; cancer stem cells; cancer therapy; hematopoiesis; leukemia stem cells; myelofibrosis; myeloproliferative neoplasms; secondary AML; splicing.

Figure 1.. Quantification of ADAR1p150 by Splice Isoform RNA Sequencing (RNA-seq)

(A) RNA-seq-based quantification (counts per million, CPM) of ADAR-201 (GRCh38 ENST00000368471.8, ADAR1 p110-encoding), ADAR-202 (ENST00000368474.9, ADAR1 p150-encoding), and ADAR-208 (ENST00000529168.2, ADAR1 p150-encoding 3’UTR truncated transcripts) was performed on FACS-purified hematopoietic stem cells (HSC, CD34⁺CD38⁻Lin⁻) from young (YBM; n=4) and aged bone marrow (ABM; n=4) HSC, polycythemia vera (PV, n=3), essential thrombocythemia (ET, n=2), myelofibrosis (MF, n=24), chronic myeloid leukemia (CML, n=5), or secondary acute myeloid leukemia (sAML, n=5). RNA-seq analyses were also performed on FACS-purified hematopoietic progenitor cells (HPC, CD34⁺CD38⁺Lin⁻) from primary samples, including YBM (n=8), ABM (n=8), PV (n=6), ET (n=2), MF (n=24), CML (n=5), de novo (dnAML and sAML) (n=13) AML. Statistics for HSC: ADAR-201 p<0.05 for MF, CML, and sAML versus ABM; ADAR-202 p<0.05 for MF versus ABM; ADAR-208 differences were not significant in HSC. Statistics for HPC: ADAR-201 p<0.05 for PV, ET, MF, and CML versus ABM; ADAR-202 p<0.05 for PV, ET, MF, and CML versus ABM; ADAR-208 p<0.05 for PV, ET, and MF versus ABM. Statistical analyses were performed using Student’s t-tests comparing MPNs and sAML versus ABM. (B) Structural diagram showing the spliceosome core complex with Rebecsinib interacting at the interface of SF3B1 and PHF5A, adapted from the spliceosome complex bound to pladienolide B. (C) Schematic diagram of the primary ADAR1 p150-encoding transcript, ADAR-202, and proposed Rebecsinib-induced intron retention reducing transcript expression after treatment.

Figure 2.. Development of a lentiviral ADAR1 A-to-I RNA editing reporter

(A) Schematic diagram demonstrating the synthetic RNA sequence containing an ADAR1-sensitive stop codon that, upon A-to-I editing, reads through to produce nanoluciferase and GFP proteins separated by a T2A cleavage site. (B) ADAR protein expression levels in 293T cells co-transfected with the ADAR1 nanoluciferase-GFP (nanoluc-GFP) reporter and increasing amounts of FLAG-tagged wild-type (WT) ADAR1, catalytically inactive mutant ADAR1 (E912A), or wild-type ADAR2. β-actin was used as a loading control. (C) Relative luciferase signals in 293T cells prepared as in panel B. Data are represented as mean ± SEM. (D) Live cell fluorescent imaging of GFP expression in human myeloid leukemia TF-1a cells transduced with the ADAR1 nanoluc-GFP reporter vector (lower panels) compared to untransduced controls (upper panels). (E) Detection of nanoluciferase expression via in vivo bioluminescence (IVIS) imaging of no transplant control (far left), K562-nanoluc-GFP and pCDH vector transduced and K562-nanoluc-GFP and ADAR1 wild-type or E912A mutant transduced human leukemia cells (K562) transplanted into RAG2^−/−γc^−/− mice. (F) Luminescence-based quantification of ADAR1-dependent nanoluciferase signals in CD34⁺ cells from primary, high-risk MF samples (*=untreated patient) after in vitro transduction with the ADAR1 nanoluc-GFP reporter and treatment with vehicle control (DMSO) or Rebecsinib (72 hr). Relative luciferase signals were normalized to cell viability for each condition. Data are represented as mean ± SEM. p<0.05 compared to DMSO controls by pairwise t-test. (G) Intracellular flow cytometry-based quantification of STAT3 phosphorylation (expressed as mean fluorescence intensity, MFI, values within HPC populations) after in vitro treatment with vehicle control (DMSO) or Rebecsinib (1 μM, 72 hr). See also Figure S1.

Figure 3.. Rebecsinib Inhibits ADAR1p150 mediated high-risk MF HPC and LSC survival

(A) Schematic diagram of in vitro MF HPC and LSC survival and self-renewal assays. (B) Flow cytometry-based viable cell counts (5,000 events measured) in high-risk MF samples after in vitro treatment of primary CD34⁺ cells with vehicle control (DMSO) or Rebecsinib (72 hr). (C) Flow cytometry-based quantification of ADAR1 p150 protein expression in high-risk MF samples after in vitro transduction of primary CD34⁺ cells with ADAR1 nanoluc-GFP reporter or vector control (pCDH) followed by treatment with vehicle control (DMSO) or Rebecsinib (72 hr). (D, E) Quantification of colony formation (survival, D) and replating (self-renewal, E) of high-risk MF HPC and sAML LSC compared with cord blood (CB) and aged versus young normal bone marrow (a-NBM, y-NBM) controls treated with Rebecsinib at increasing concentrations. Bar graphs show data as mean ± SEM and statistical analyses by pairwise t-test and dose-response assays show data as mean ± SD and statistical analyses by one-way ANOVA. See also Figure S1.

Figure 4.. Rebecsinib pharmacodynamic and pharmacokinetic studies in pre-clinical and pre-IND models

(A) Quantification of cell viability (left panel) and splicing modulation (RFP/GFP ratios, right panel) by flow cytometry analyses of the human AML cell line (KG-1a) stably transduced with a lentiviral dual-fluorescence splicing reporter vector and treated with increasing concentrations of Rebecsinib. (B) Transcript diagram illustrating alternative splicing of MCL1 to generate MCL1-short (S, pro-apoptosis) and MCL1-long (L, anti-apoptosis) variants. For human cells, the ratio of MCL1-short to long isoforms is shown. MCL1 S/L ratios in Rebecsinib dose response assays performed using primary sAML LSC (without stromal co-culture). Splice isoform-specific qRT-PCR values were normalized to DMSO-treated controls for each individual patient sample. (D) Schematic diagram outlining multispecies toxicokinetic (TK) and pharmacodynamic studies in mammalian species treated in vivo with a single dose of Rebecsinib. Toxicokinetic analyses were performed in rats (n=6 per sex, per group), rabbits (n=2 per sex, per group), and non-human primates (NHPs, n=1 per sex, per group) and pharmacodynamic splice isoform quantification studies were performed in peripheral blood mononuclear cells (PBMCs) from NHP. (E, F) For toxicokinetic analyses, rats (E) and rabbits (F) were given a single injection of Rebecsinib at 1-40 mg/kg, or vehicle control, and blood samples were drawn at regular intervals to determine plasma concentrations of the compound over 8 hr after treatment. (G, H) For in vivo TK and complementary pharmacodynamic studies, NHPs were given a single injection of Rebecsinib at 3-20 mg/kg, or vehicle control, and blood samples were drawn at regular intervals to determine plasma concentrations (G) of the compound along with splice isoform biomarker assays to quantify MCL1 exon skipping in PBMCs isolated from treated animals (H, n=2 animals per group). All data are represented as mean ± SEM. See also Figures S2 and S3.

Figure 5.. ADAR1 expression and LSC self-renewal following Rebecsinib treatment

(A) Schematic diagram showing in vivo treatment of primary patient LSC or cord blood (CB)-engrafted mice and sAML serial transplantation studies. (B) Flow cytometry analysis quantifying human LSC survival in sAML-engrafted mice treated with Rebecsinib Lot 1 (n=3 sAML50261, n=3 sAML 2008-5) or Lot 2 (n=4 sAML50261) compared with vehicle control (n=4 sAML50261, n=3 sAML2008-5) at 10 mg/kg twice weekly for two weeks (5 total doses). (C) qRT-PCR analyses in CD34+ cells isolated from the spleens of sAML50261 mice treated with Rebecsinib (as in A) showing decreased total ADAR1 expression by qRT-PCR. (D) Mean fluorescence intensity (MFI) of ADAR1p150 protein levels in human HSCs (CD45⁺CD34⁺CD38⁻Lin⁻) and HPCs (CD45⁺CD34⁺CD38⁺Lin⁻) from the spleens of sAML50261 engrafted mice treated with Rebecsinib or vehicle. (E) Whole transcriptome-based RNA editing analyses of previously-described RNA-seq data generated from CD34⁺ cells isolated from the spleens of sAML50261 engrafted mice treated with Rebecsinib or vehicle. Total edits, edits in unique genes, and normalized numbers of edits per million reads were calculated using RNA editing pipelines as previously described. (F) Isoform-level analysis of MCL1 transcripts from RNA-sequencing data shown in panel E. (G) Overall mouse survival in serially transplanted sAML50261 mice (primary transplanted mice were treated with Rebecsinib or vehicle) (n=10 mice per group). (H) Ratios of ADAR1p150-3’UTR truncated (ADAR-208) to ADAR1p110 (ADAR-201) by RNA-seq analyses of CD34⁺ cells isolated from serial transplant recipients of sAML50261 LSC engrafted mice treated with vehicle or Rebecsinib. Serial transplant recipients received no further treatment. Bar graphs show data as mean ± SEM and statistical analyses by unpaired t-test, and overall animal survival plot shows Kaplan-Meier plots (p=0.0008). See also Figures S4 and S5.