Targeting β-catenin degradation with GSK3β inhibitors induces cell death in acute lymphoblastic leukemia

Abstract

As part of canonical Wnt signaling, T cell factor (TCF)-β-catenin complexes promote MYC-dependent proliferation. Lesions of the β-catenin protein degradation machinery are common oncogenic drivers. Here, we show that B cell acute lymphoblastic leukemia (B-ALL) lacks these mutations and critically depends on unencumbered β-catenin protein degradation. Compared to solid tumors, we found that mouse and human B-ALL express β-catenin protein at much lower levels; β-catenin protein was constitutively phosphorylated by glycogen synthase kinase 3B (GSK3β) and poised for proteasomal degradation. Instead of TCF-β-catenin complexes to activate MYC, β-catenin paired with B lymphoid Ikaros and NuRD complex factors, resulting in MYC repression and acute cell death. To leverage β-catenin protein degradation as a previously unrecognized vulnerability in B-ALL, we validated GSK3β inhibition in patient-derived xenograft models in vivo. CRISPR screens confirmed β-catenin protein degradation as a central mechanistic target of established GSK3β inhibitors. As several GSK3β inhibitors achieved favorable safety profiles in clinical trials, our results provide a rationale for repurposing these compounds for persons with refractory B cell malignancies.

Fig. 1. B lymphoid cells express β-catenin mRNA but lack β-catenin protein expression.

a, Representative immunohistochemical staining for β-catenin (brown) on tissue microarrays from lung cancer (n = 15 tumors, biological replicates), colon cancer (n = 25 tumors, biological replicates) and malignant melanoma (n = 5 tumors, biological replicates) in comparison to MCL (n = 26 tumors, biological replicates), follicular lymphoma (n = 38 tumors, biological replicates), DLBCL (n = 35 tumors, biological replicates) and HD (n = 44 tumors, biological replicates). b, Western blot analysis for β-catenin, β-tubulin and TBP on nuclear fractions of lung cancer (n = 7 cell lines, biological replicates), colon cancer (n = 6 cell lines, biological replicates), malignant melanoma (n = 3 cell lines, biological replicates), B-ALL (n = 7 cell lines, biological replicates), DLBCL (n = 5 cell lines, biological replicates), MCL (n = 3 cell lines, biological replicates), Burkitt’s lymphoma (n = 3 cell lines, biological replicates), HD (n = 2 cell lines, biological replicates) and multiple myeloma (MM; n = 2 cell lines, biological replicates). c, Top, transcriptional analysis of 779 cancer cell lines by RNA-seq (PRJNA523380) for the expression of CTNNB1 in 692 solid tumor cell lines (biological replicates) compared to 87 B cell leukemia and lymphoma cell lines (biological replicates). Bottom, protein levels of β-catenin were measured by RPPA (CCLE2019) in B cell leukemia and lymphoma cell lines (n = 86 cell lines, biological replicates), as well as solid tumor cell lines (n = 646 cell lines, biological replicates). TPM, transcripts per million. d,e, Fluorescent protein stability assay to compare rate of β-catenin degradation in epithelial cell lines, including colon (SW480) and lung (H82) cancer cell lines and cells derived from normal epithelial tissues including human embryonic kidney (HEK) and normal breast epithelial cells (MCF10) and in B lymphoid cells including B-ALL (KOPN8 and TOM1), MCL (JEKO) and DLBCL (TMD8). GFP-tagged WT β-catenin (protein) or GFP tag was expressed along with mScarlet (mRNA) reporter and stability of β-catenin was assessed by flow cytometric measurement of GFP and mScarlet signal. d, Representative FACS plots are shown for cells expressing β-catenin–GFP fusion protein along with mScarlet mRNA reporter. e, Frequencies of cells expressing β-catenin–GFP fusion protein within mScarlet⁺ cells are shown and an unpaired two-tailed t-test was performed to compare β-catenin protein and mRNA expression in all B lymphoid cells (B-ALL and B-NHL; n = 10 cell lines, biological replicates) versus all epithelial cells (normal epithelial, colon cancer and lung cancer; n = 9 cell lines, biological replicates) (P = 2.23 × 10⁻⁸). f, Representative immunohistochemical staining for β-catenin (brown; top) and β-catenin phosphorylated by GSK3β (S37, brown; bottom) on tissue microarrays from lung cancer (n = 41 tumors, biological replicates), colon cancer (n = 13 tumors, biological replicates), breast cancer (n = 58 tumors, biological replicates), ovarian cancer (n = 9 tumors, biological replicates), malignant melanoma (n = 13 tumors, biological replicates), MCL (n = 21 tumors, biological replicates), follicular lymphoma (n = 54 tumors, biological replicates), DLBCL (n = 47 tumors, biological replicates) and HD (n = 20 tumors, biological replicates) using hematoxylin (blue) as the counterstain. g, Western blot analysis for total β-catenin, N-terminal phosphorylated β-catenin by GSK3β (S33, S37 and T41) and nonphosphorylated (active) β-catenin in a panel of lung (n = 2 cell lines, biological replicates) and colon (n = 4 cell lines, biological replicates) cancer cell lines in comparison to human B-ALL (n = 3 cell lines, biological replicates), MCL (n = 2 cell lines, biological replicates) and postgerminal center lymphoma (n = 2 cell lines, biological replicates) (three technical repeats). The exposure time for total and nonphosphorylated β-catenin was 185 s; the exposure time for N-terminal phosphorylated β-catenin (S33, S37 and T41) was 1,199 s. Source data

Fig. 2. B cell malignancies are uniquely dependent on efficient β-catenin protein degradation.

a, Frequencies of nonsynonymous coding mutations of β-catenin (CTNNB1; filtered for hotspot mutations of GSK3β and CK1α phosphorylation sites), APC, AXIN1, AXIN2, CSNK1A and GSK3B are shown for 14 types of cancer including B cell leukemia and lymphoma (n = 4,387 tumors, biological replicates) and solid tumors (n = 86,426 tumors, biological replicates). b,c, Apc^fl/+ pre-B cells were transformed with BCR–ABL1 or NRAS^G12D to establish B-ALL cell lines and subsequently transduced with vectors expressing 4-OHT-inducible Cre (Cre-ER^T2) or EV control (ER^T2) along with GFP. b, Changes in frequencies of GFP⁺ cells were studied by FACS for 0–8 days after 4-OHT addition. Representative data from two independent experiments are shown (n = 3 biological replicates). Data are presented as the mean ± s.d. c, Apc^fl/+ B-ALL cells expressing Cre-ER^T2 or ER^T2 constructs (sorted for GFP⁺) were plated on methylcellulose 1 day after 4-OHT treatment and colonies were counted 14 days after plating. Colony numbers for cells (ER^T2 versus Cre-ER^T2) carrying NRAS^G12D: 680 ± 61 versus 16 ± 8. Colony numbers for cells (ER^T2 versus Cre-ER^T2) carrying BCR–ABL1: 94 ± 22 versus 2 ± 2. Data are shown as the mean ± s.d. (n = 3 biological replicates, two independent experiments). d,e, BCR–ABL1 or NRAS^G12D transformed Gsk3b^fl/+ B-ALL cells were transduced with constructs expressing Cre-ER^T2 or ER^T2 and GFP. d, Increases or decreases in frequencies in GFP⁺ cells were studied by FACS for 0–8 days after 4-OHT addition. Data show the mean ± s.d. calculated from three independent experiments (n = 3 technical replicates). e, GFP⁺ B-ALL cells were plated for colony formation assays 1 day after 4-OHT treatment. Representative images from two independent experiments are shown for 14 days after plating (colony numbers for ER^T2 versus Cre-ER^T2 cells (mean ± s.d.) with NRAS^G12D: 68 ± 16 versus 1 ± 1; colony numbers for cells (ER^T2 versus Cre-ER^T2) carrying BCR–ABL1: 261 ± 47 versus 3 ± 2) (n = 3 technical replicates). f,g, BCR–ABL1 or NRAS^G12D transformed cells with monoallelic CK1α deletion (Ck1a^fl/+; Cre-ER^T2) and CK1α WT cells (Ck1a^fl/+; ER^T2). Changes in frequencies of GFP⁺ cells were measured by FACS for 8 days. Data are representative of two independent experiments (n = 3 biological replicates) and presented as the mean ± s.d. g, Representative images and average number of colonies 10 days after plating (two independent experiments). h,i, β-catenin^{(S33;S45)+/fl} B-ALL cells (BCR–ABL1) were transduced with vectors expressing GFP-tagged Cre-ER^T2(Cre) or ER^T2(EV). GFP⁺ cells were sorted by FACS and injected into sublethally irradiated NSG mice 2 days after 4-OHT treatment. ELDA was performed to assess effects of β-catenin accumulation on leukemia-initiation capacity (LIC) of B-ALL cells. h, Kaplan–Meier analysis performed for calculating survival in each group. Log-rank test for survival of 120,000 Cre versus 120,000 EV transplanted mice (n = 4 mice in each group, biological replicates), P = 0.0067. Log-rank test for 6,000 Cre versus 6,000 EV transplanted mice (n = 4 mice each group, biological replicates), P = 0.0067. Log-rank test for 300 Cre (n = 4 mice, biological replicates) versus 300 EV transplanted mice (n = 5 mice, biological replicates), P = 0.18. i, LIC was determined in B-ALL cells with β-catenin accumulation (1 in 40,063 cells) and control cells (1 in 1,042) using Chi-square test and 90% CI is depicted (P = 0.0006). (j,k) β-catenin^{(S33-S45)+/fl} B-ALL (BCR–ABL1) cells were transduced with doxycycline-inducible vectors expressing myeloid transcription factor CEBPα or EV and were subsequently transduced with GFP-tagged Cre-ER^T2 or ER^T2 vectors for excision of β-catenin GSK3β and CK1α phosphorylation sites. CEBPα-driven myeloid reprogramming was induced by addition of doxycycline. (j) Flow cytometry analysis to identify myeloid (CD11b⁺) and B lymphoid (CD19⁺) cells two days after doxycycline treatment. k, Western blot analysis to measure CEBPα, Ikzf1, Ikzf3 and Myc levels following β-catenin accumulation in B-ALL after CEBPα myeloid reprogramming (CEBPα) or EV conditions (n = 3 independent repeats). l, Changes in frequencies of GFP⁺ cells were monitored by FACS for 0–6 days after 4-OHT-mediated activation of Cre and accumulation of β-catenin. Data are shown as the mean ± s.d. from three independent experiments (n = 3 technical replicates). Source data

Fig. 3. Impaired β-catenin protein degradation in B cells results in transcriptional repression of MYC.

a,b, BCR–ABL1 transformed β-catenin^{(S33;S45)+/fl} B-ALL cells expressing Cre-ER^T2 or ER^T2 vectors were treated with 4-OHT to excise GSK3β and CK1α phosphorylation sites on β-catenin. Gene expression changes were studied by RNA-seq (GSE305472) 1 day after 4-OHT-mediated β-catenin stabilization (n = 4 cell lines, biological replicates). a, GSEA identified depletion of Myc target genes and enrichment of Wnt–β-catenin signaling as top-ranking gene sets following β-catenin accumulation. b, Genes that were upregulated (n = 354) or downregulated (n = 119) upon β-catenin stabilization are shown as a heat map. Genes with the most prominent upregulation included Prdm1 and negative regulators of Wnt–β-catenin signaling such as Axin2, Kremen1, Nkd1, Nkd2 and Tle3, while Myc and Pax5 were amongst downregulated genes. c, Western blot analysis to detect changes in Prdm1 and Myc levels following β-catenin stabilization in β-catenin^{(S33;S45)+/fl} B-ALL cells using β-actin as loading control (n = 2 independent experiments). d, Pre-B cells from transgenic mice expressing a dual Myc-β-catenin reporter were transformed with BCR–ABL1 to develop a B-ALL cell line. Dual-reporter cells with expression of an N-terminal eGFP–Myc fusion and N-terminal mTurquoise–β-catenin fusion protein (Myc–eGFP × mTq–β-catenin) were treated with 10 nM LY2090314 for 12 h. Changes in Myc–eGFP and mTq–β-catenin protein expression were assessed by measuring eGFP and mTurquoise signal, respectively. Representative confocal microscopy images (top) for eGFP–Myc (green) and mTq–β-catenin (red) are shown for 0–12 h of LY2090314 treatment. FACS plots (bottom) for measuring mTq–β-catenin (x axis) and eGFP–Myc (y axis) levels are presented for WT and Myc–β-catenin reporter cells for the indicated time points of LY2090314 treatment. FACS and time-lapse imaging experiments were performed two times each. e–g, BCR–ABL1 transformed β-catenin^{(S33;S45)+/fl} B-ALL cells expressing Cre-ER^T2 or ER^T2 (puromycin selected) were transduced with GFP-tagged Myc or EV. e, Confirmation of β-catenin and Myc expression in FACS-sorted GFP⁺ cells by western blot 3 days after 4-OHT treatment. f, FACS analysis to monitor enrichment or depletion of GFP⁺ cells (Myc versus EV) upon β-catenin activation. Data are presented as the mean ± s.d. calculated from three independent experiments (n = 3 technical replicates). g, Colony formation abilities of cells expressing Myc or EV were assessed 2 days after 4-OHT-induced β-catenin accumulation. Data shown are representative of two independent experiments (n = 3 technical replicates). Colony numbers for each condition: ER^T2 EV, 208 ± 32; Cre-ER^T2 EV, 21 ± 7; ER^T2 Myc, 66 ± 36, Cre-ER^T2 Myc, 93 ± 36 (mean ± s.d). h, BCR–ABL1 transformed B-ALL cells from Ctnnb1^fl/fl mice were transduced with vectors expressing Cre-ER^T2 or ER^T2 and treated with 4-OHT for 2 days to induce β-catenin deletion. Western blot was performed to analyze β-catenin and Myc levels 16 h after LY2090314 treatment (40 nM; middle) in β-catenin knockout (Ctnnb1^−/−, Cre-ER^T2) or WT (Ctnnb1^+/+, ER^T2) B-ALL cells (n = 3 independent experiments). i, Human B-ALL cells (PDX2) were edited with HDR to generate cells with homozygous or heterozygous deletion of CTNNB1. β-catenin-deficient (gCTNNB1) or WT (gNT) PDX2 cells were treated with LY2090314 (20 nM) for 16 h. Western blot was performed to analyze β-catenin and MYC levels following GSK3β inhibition (n = 2 independent experiments). j, β-catenin knockout (Cre-ER^T2) or WT (ER^T2) B-ALL cells were treated with LY2090314 for 16 h and RNA-seq was performed to characterize transcriptomic changes (GSE245287). Principal component (PC) analysis of gene expression changes in Ctnnb1^fl/fl B-ALL cells upon β-catenin deletion (Cre-ER^T2) compared to WT (ER^T2) cells (n = 3 mice, biological replicates) in the presence or absence of GSK3β inhibition. Although GSK3β inhibition in β-catenin competent cells induced major changes in gene expression (ER^T2 + LY2090314 versus ER^T2), this effect was reversed by β-catenin deletion (Cre-ER^T2 + LY2090314). Source data

Fig. 4. β-catenin forms repressive complexes with B lymphoid transcription factors Ikzf1 and Ikzf3.

a, Proteins bound to β-catenin in BCR–ABL1 transformed B-ALL cells from β-catenin^{(S33;S45)+/fl} mice were enriched by co-IP, identified by mass spectrometry and plotted on the basis of statistical significance and log₂ fold enrichment over IgG background control (n = 4 technical replicates). Proteins with the most prominent binding to β-catenin included Ikaros factors Ikzf1 and Ikzf3 (red) and members of the repressive NuRD complex Chd4, Gatad2a, Gatad2b, Mta1, Mta2, Mdb3, Rbbp4, Hdac1 and Hdac2 (blue). Linear modeling and empirical Bayes testing used for differentially enriched proteins. b, β-catenin-interacting proteins were validated by co-IP and western blot in whole-cell lysates (input), proteins bound (elute) and flowthrough (FT) to isotype control or antibodies against β-catenin, using Stat5 as a negative control (n = 2 independent experiments). c, Human B-ALL (MXP2), B cell lymphoma (JEKO1), AML (MOLM13), colon cancer (SW480) and lung cancer (H446) cell lines were engineered to express doxycycline-inducible degradation resistant form of β-catenin. Co-IP experiments with antibodies to β-catenin or control Ig were performed 1 day after doxycycline treatment. Eluted proteins were analyzed by mass spectrometry. The heat map shows the interaction score (y axis, log₂ fold change) of β-catenin-binding proteins normalized to Ig control in B-ALL, MCL, AML, colon cancer and lung cancer cell lines. FC, fold change. d, Whole-cell lysates (input), proteins bound (elute) and FT with β-catenin antibodies or control Ig were analyzed by western blotting to study interactions between β-catenin and Ikaros factors (IKZF1 and IKZF3), NuRD complex components (MTA1, MTA2 and GATAD2A) and TCF7L2 in B-ALL (PDX2), myeloid leukemia (JURL-MK1) and colon cancer (SW620) cells 16 h after pharmacological β-catenin stabilization (LY2090314, 20 nM) (n = 2 independent experiments). e–i, BCR–ABL1 transformed β-catenin^{(S33;S45)+/fl} B-ALL cells were gene-edited with gRNAs targeting Ikaros factors (Ikzf1 and Ikzf3; individually or both) or gNT. Deletion of Ikaros factors was confirmed by western blot in clonal cell lines established from single cells. Multiple clones were studied for each genotype. β-catenin^{(S33;S45)+/fl} B-ALL cells were transduced with 4-OHT-inducible GFP-tagged Cre-ER^T2 or ER^T2 constructs. e, Changes in β-catenin interactomes in B-ALL cells upon Ikaros factor deletion (gIkzf1/3) were analyzed by co-IP and mass spectrometry. Proteins bound to β-catenin were plotted on the basis of significance (−log₁₀ P value; y axis) and abundance change (log₂ fold enrichment; x axis) compared to B-ALL cells without deletion of Ikaros factors (gNT; n = 3 technical replicates). DKO, double knockout. f, GSEA plots show the depletion of MYC target genes upon β-catenin accumulation in the presence of Ikaros factors (top), which were reversed by the deletion of Ikaros factors (bottom) (GSE196767). g, Colony-forming assays for B-ALL cells with (Ikzf1/3 double knockout) and without (Ikzf1/3 WT) Ikaros factor deletion were performed upon β-catenin stabilization (Cre-ER^T2) compared to baseline β-catenin expression (ER^T2). Representative images and mean colony numbers from three independent experiments each with three technical replicates are shown 10 days after plating. h, Western blot analysis to study changes in Myc levels in single-cell-derived B-ALL clones with deletion of Ikzf1 and/or Ikzf3 (white box) in comparison to clones WT for both Ikaros factors (dark-green box) following β-catenin stabilization (red box) compared to baseline conditions (light-green box) (n = 3 independent experiments). i, The competitive fitness of B-ALL clones was assessed for baseline β-catenin levels (ER^T2, light-green box) or following β-catenin accumulation (Cre-ER^T2, red box) in Ikzf1 and Ikzf3 WT cells (gNT, dark-green box) or upon deletion of Ikzf1 and/or Ikzf3 (gIkzf1, gIkzf3; white box). Data are presented as the mean ± s.d. of three independent experiments (n = 3 technical replicates). Source data

Fig. 5. Identification of an Ikaros motif in the MYC BENC enhancer region required for β-catenin-mediated repression of MYC.

a–e, BCR–ABL1 transformed β-catenin^{(S33;S45)+/fl} B-ALL cells were edited with gIkzf1/3 or gNT. Clonal cell lines that are WT or knockout for Ikzf1/3 were generated and transduced with 4-OHT-inducible Cre-ER^T2 or ER^T2 vectors. Transcriptional targets of Ikzf1, Ikzf3 and β-catenin and changes in histone mark H3K27ac were studied by ChIP-seq in Ikzf1/3 WT (dark-green box) or Ikzf1/3 knockout (white box) cells in the presence (red box) or absence (light-green box) of β-catenin stabilization (GSE196745). a, Venn diagram showing the number of regions bound by β-catenin only (4,356), Ikaros factors only (4,596) or both (11,354). Of 15,710 β-catenin peaks, 11,354 (72.2%) were also bound by Ikaros factors. b, ChIP-seq analysis of histone mark H3K27ac is shown as a heat map for the Myc locus, including upstream Myc promoter regions and long-range transcriptional enhancers of Myc in B-ALL cells from β-catenin^{(S33;S45)+/fl} mice. H3K27ac distribution marking active enhancer regions shows that most of the H3K27ac enhancer activity is concentrated in BENC regions in B-ALL cells. In humans, single-nucleotide polymorphisms in this region are associated with increased risk for B-ALL (rs4617118, rs75777619 and rs28665337). c, Enrichment of β-catenin, Ikzf1 and Ikzf3 binding to BENC elements C and D and changes in H3K27ac upon induction of β-catenin (red box) in B-ALL cells with (white box) or without (dark-green box) deletion of Ikaros factors compared to baseline conditions (light-green box). Identification of Ikaros-binding motifs in the BENC-C (m1 and m2) and BENC-D (m3) elements. d, Quantification of H3K27ac signals (mean value of n = 2 cell lines, technical replicates) by ChIP-seq at BENC enhancer regions, other regions with binding of both Ikaros factors and β-catenin (cobound) and all other regions. H3K27ac ChIP was performed with (red box) and without (light-green box) β-catenin accumulation and in the presence (white box) or absence (dark-green box) of Ikaros factor deletion. Box plots display boxes at the first and third quartiles with a line at the median and whiskers from the minimum to maximum data points within 1.5× the interquartile range (BENC, n = 2; cobound, n = 1,043; other, n = 42 peaks). e, ChIP–qPCR analysis for recruitment of NuRD complex components (MTA2 and CHD4) to the BENC-C enhancer region in β-catenin^{(S33;S45)+/fl} B-ALL cells that are WT (dark-green box) or knockout (white box) for Ikzf1/3 under baseline β-catenin (light-green box), following β-catenin stabilization (red box) or deletion of β-catenin (white box). Data represent a pool of six independent experiments. Mean values from independent experiments (n = 3 technical replicates) were used to compare changes in MTA and CDH4 binding upon β-catenin stabilization compared to baseline and β-catenin knockout condition compared to baseline (two-sided unpaired t-test). f, HDR-mediated editing of the BENC-C m1 motif to abrogate binding of Ikzf1 and Ikzf3 by mutating Ikaros core motif GGGAA and generation of an EcoR1 recognition sequence (GAATTC). Following EcoR1 digestion to confirm mutation of the BENC-C m1 motif, single-cell-derived clones were generated and clones carrying the BENC-C m1 motif mutation were identified by Sanger sequencing. g,h, Single-cell-derived clones from β-catenin^{(S33;S45)+/fl} B-ALL cells with WT or mutated BENC-C Ikaros m1 motifs were transduced with vectors expressing Cre-ER^T2 or ER^T2 along with GFP. g, GFP⁺ cells were sorted by FACS and western blot analysis was performed to characterize β-catenin and Myc protein levels 1 day after 4-OHT treatment in B-ALL cells carrying intact or mutated BENC-C Ikaros m1 motifs in the presence (Cre-ER^T2) or absence (ER^T2) of β-catenin stabilization. h, GFP⁺ cells with intact (WT 1G11) or mutated BENC-C Ikaros m1 motifs (mut 1C12) were mixed with cells with an intact BENC-C Ikaros m1 motif and the effects of β-catenin stabilization were measured in competitive cell culture experiments by flow cytometry. Data depict the means ± s.d. calculated from three independent experiments (n = 3 technical replicates). Source data

Fig. 6. Small-molecule inhibition of β-catenin protein degradation induces B cell-selective cell death.

a, Meta-analysis of cell-type-specific toxicities across three compound screens (CTD2, GDSC1 and GDSC2). The ΔAAC compares drug sensitivity in B cell lines (B-ALL, MCL and DLBCL) to solid tumor cell lines using the Wilcoxon effect size test. Inhibitors of GSK3β (AZD7969, GSK3iIX, CHIR99021 and ML320) show increased sensitivity in B lymphoid compared to solid tumors. b, Compound set enrichment analysis for GSK3β inhibitors was ranked by differential ΔAAC effect sizes as shown in a and demonstrated enrichment for B cell-selective effects (q = 0.008, NES = 1.78). c, Sensitivity to GSK3β inhibitors (ΔAAC; x axis) and β-catenin protein levels as measured by RPPA (CCLE2019; y axis) were plotted for human B lymphoid (n = 76 cell lines, biological replicates; red circles) and solid (n = 646 cell lines, biological replicates; gray circles) tumor cell lines. d, Human B-ALL (PDX2) cells were treated for 0–8 h with 10 nM LY2090314. Changes in total β-catenin, N-terminal phosphorylated β-catenin (S33, S37 and T41) and active β-catenin (nonphosphorylated) levels were analyzed by western blot. The exposure time for total and nonphosphorylated β-catenin was 185 s, whereas the exposure time for phosphorylated β-catenin (S33, S37, T41) was 1,199 s (n = 2 independent repeats). e, Drug responses to GSK3β inhibitor LY2090314 (EC₅₀ values, nM) were calculated in B cell tumors (B-ALL, CLL and MCL; n = 34 cell lines, biological replicates; red circles), by measuring luminescence on day 3 and fitting of three-parameter log-logistic dose–response curves and comparing to epithelial cancer cell lines (n = 343 cell lines, biological replicates; Prism drug-repurposing screen). Data are presented as the mean ± s.d. f, B-ALL (MXP2, LAX2, BLQ5 and IAH8R), MCL (Z138), colon cancer (SW480, SW620 and LOVO) and lung cancer (H446 and H82) cell lines were treated with the GSK3β small-molecule inhibitor LY2090314 (20 nM) for 1 day. β-catenin, MYC, IKZF1 and IKZF3 protein levels were assessed by western blot, using β-actin as a loading control (n = 3 independent experiments). g, Computational analyses of correlations between gene expression (biomarker) and sensitivity to the GSK3β inhibitor CHIR99021 in B lymphoid and epithelial tumor cell lines are shown as a volcano plot with positive and negative correlation coefficients (x axis) and statistical significance (−log₁₀ q value; y axis). B cell-specific expression of IKZF1 and IKZF3 was strongly correlated with high sensitivity to CHIR99021, while epithelial-specific TCF7L1 and TCF7L2 expression correlated with CHIR99021 resistance. h, To comprehensively identify mechanistic targets of GSK3β inhibition in human B-ALL cells, we performed a chemogenomic CRISPR screen. NALM6 B-ALL cells bearing an integrated inducible Cas9 expression cassette were transduced with the genome-wide knockout EKO sgRNA library (278,754 sgRNAs, targeting 22,956 genes, with 12 sgRNAs per gene and negative controls) and treated with LY2090314 at 3.5 nM to induce partial GSK3β inhibition for 8 days. Context-dependent chemogenomic interaction scores and P values were calculated using a modified version of the RANKS algorithm, which uses sgRNAs targeting similarly essential genes as controls to distinguish condition-specific chemogenomic interactions from nonspecific fitness and essentiality phenotypes. i, Pre-B cells from β-catenin WT (Mb1-Cre × Ctnnb1^+/+) and knockout (Mb1-Cre × Ctnnb1^fl/fl) were transformed with BCR–ABL1 to establish B-ALL cell lines. LY2090314 sensitivity was evaluated in β-catenin WT (Mb1-Cre × Ctnnb1^+/+) and knockout (Mb1-Cre × Ctnnb1^fl/fl) B-ALL cells by measuring luminescence 3 days after LY2090314 treatment (0–100 nM). Changes in viability were calculated by normalizing luminescence signals from treated cells to baseline values of untreated cells. Data are presented as the mean ± s.d. of three independent experiments (n = 3 technical replicates). j, β-catenin deletion was introduced in human B-ALL cell lines (BV173) and PDX (PDX2) cells by CRISPR–Cas9 using gCTNNB1 or gNT. Human B-ALL cells that are WT (gNT) or knockout for β-catenin (gCTNNB1) were treated with LY2090314 and viability changes were assessed by measuring luminescence 3 days after treatment. Data are presented as the mean ± s.d. of three independent experiments (n = 3 technical replicates). Source data

Fig. 7. Rationale for repurposing of GSK3β inhibitors for refractory B lymphoid malignancies.

a, Sensitivity to the GSK3β inhibitor LY2090314 was assessed in a panel of lymphoid (B-ALL, CLL, MCL and postgerminal center (post-GC) lymphoma; n = 18 cell lines, biological replicates), myeloid (CML and AML; n = 4 cell lines, biological replicates) and epithelial (colon and lung cancer; n = 7 cell lines, biological replicates) cancer cell lines by measuring luminescence signal 3 days after treatment. Growth-inhibitory effects are shown as a heat map. Asterisks denote B cell lines with MYC translocation. b, Drug responses (area under the curve) to the GSK3β inhibitor CHIR99021 (CTD2 screen) were plotted for 84 lymphoma and leukemia cell lines, comparing cell lines with (8q24⁺; n = 31 cell lines, biological replicates) and without (8q24⁻; n = 53 cell lines, biological replicates) MYC rearrangement as determined by fluorescence in situ hybridization or conventional cytogenetics. A two-sided unpaired t-test was used to compare groups. Data are presented as the mean ± s.d. c, Cell viability measurements following LY2090314 treatment were compared for human B-ALL samples from participants who responded to conventional chemotherapy (sensitive) and from participants with acquired chemoresistance (refractory). Boxes below the heat map indicate B-ALL samples with IKZF1 lesions (deletion or mutation; red box). d,e, Luciferase-labeled LAX2 B-ALL cells (relapse sample) were injected into sublethally irradiated NSG mice. Mice were either treated with 12.5 mg kg⁻¹ LY2090314 or vehicle control twice daily for four consecutive days per week (eight times a week) for 4 weeks. d, Leukemia burden was assessed by bioluminescence imaging on day 18 (top) and day 28 (bottom) following transplantation. e, Kaplan–Meier analysis of overall survival in each group (P = 1.2 × 10⁻⁵, calculated by log-rank test comparing vehicle (n = 10 mice, biological replicates) to LY2090314 (n = 9 mice, biological replicates)). f, Refractory B-ALL (BLQ5) cells were transplanted into sublethally irradiated NSG mice. Mice were either treated with 12.5 mg kg⁻¹ LY2090314 or vehicle control twice daily for four consecutive days a week, for 4 weeks. Kaplan–Meier analysis was performed to calculate overall survival between groups treated with vehicle (n = 8 mice, biological replicates) and LY2090314 (n = 7 mice, biological replicates) (P = 0.004). g,h, Luciferase-labeled PDX2 cells (diagnosis sample) were injected into sublethally irradiated NSG mice, either treated with 12.5 mg kg⁻¹ LY2090314 or vehicle control twice daily for 4 weeks. g, Leukemia burden was assessed by bioluminescence imaging on day 18 (top) and day 42 (bottom) after transplantation. h, Kaplan–Meier analysis of overall survival comparing vehicle-treated (n = 9 mice, biological replicates) to LY2090314-treated (n = 9 mice, biological replicates) group (P = 6.5 × 10⁻⁵, calculated by log-rank test). i, Refractory MCL cells (BOS4) were injected into sublethally irradiated NSG mice. Mice were either treated with 12.5 mg kg⁻¹ LY2090314 or vehicle control twice daily, for four days a week, over a period of 3 weeks. Kaplan–Meier analysis was performed to calculate overall survival in vehicle-treated (n = 9 mice, biological replicates) versus LY2090314-treated (n = 7 mice, biological replicates) group (P = 6.5 × 10⁻⁴). Source data

Extended Data Fig. 1. Targeting high-efficiency β-catenin protein degradation in B-cell leukemia.

(a) In epithelial cancers (left) efficiency of β-catenin protein degradation is low and further reduced by oncogenic mutations. β-catenin accumulates and pairs with TCF factors for transcriptional activation of MYC. In B-cell leukemia (middle) β-catenin is constitutively phosphorylated by GSK3B and targeted for B-cell-specific high-efficiency protein degradation. GSK3B-inhibitors impair β-catenin protein degradation (right) and β-catenin accumulates in the nucleus. Instead of TCF, β-catenin pairs with Ikaros and NuRD factors for MYC repression (right). b) Summary of Phase I and Phase II clinical trials with GSK3β inhibitors for a variety of clinical indications. In a total of 20 clinical trials, all tested small molecule inhibitors achieved favorable safety and PK/PD profiles at micromolar plasma concentrations (Cmax). None of the inhibitors achieved clinical responses. Moderate adverse effects included diarrhea, anemia and lymphopenia.

Extended Data Fig. 2. Lack of β-catenin expression and activity in B-lymphoid cells.

(a-b) Immunohistochemical staining for β-catenin expression (brown, hematoxylin counterstain) in (a) human lymphoid tissues (left), including bone marrow (n = 7 tissues, biological replicates), spleen (n = 6 tissues, biological replicates), lymph node (n = 8 tissues, biological replicates), tonsil (n = 9 tissues, biological replicates) and (b) epithelial tissues (right), including colon (n = 7 tissues, biological replicates), liver (n = 9 tissues, biological replicates), pancreas (n = 12 tissues, biological replicates), kidney (n = 8 tissues, biological replicates), lung (n = 8 tissues, biological replicates) and skin (n = 7 tissues, biological replicates). Representative images are shown. Data derived from: https://www.proteinatlas.org/ENSG00000168036-CTNNB1/tissue#expression_summary (c-d) β-catenin expression was assessed by immunohistochemistry (brown) and using hematoxylin as counterstain (blue) on tissue microarrays from (c) B-lymphoid malignancies, including mantle cell lymphoma (n = 12 tumors, biological replicates), follicular lymphoma (n = 24 tumors, biological replicates), DLBCL (n = 24 tumors, biological replicates) and Hodgkin’s lymphoma (n = 24 tumors, biological replicates) as well as (d) epithelial cancers, including colon cancer (n = 12 tumors, biological replicates), lung cancer (n = 12 tumors, biological replicates) and malignant melanoma (n = 3 tumors, biological replicates). (e) Western blot analysis of β-catenin expression in cytoplasmic fractions of epithelial cancers, including lung and colon cancer, malignant melanoma, as well as B-lymphoid malignancies, including B-ALL, diffuse large B-cell lymphoma (DLBCL), mantle cell lymphoma (MCL), Burkitt’s, Hodgkin’s disease (HD) and multiple myeloma cell lines. β-tubulin and TBP were used to verify purity of cytoplasmic fractions. Source data

Extended Data Fig. 3. β-catenin is constitutively phosphorylated and poised for degradation in B-lymphoid cells.

Non-phosphorylated β-catenin and β-catenin phosphorylated on N-terminal by GSK3B (pS37) were visualized by immunohistochemistry (brown, counterstain hematoxylin) on tissue microarrays from epithelial cancers, including colon cancer (n = 13 tumors, biological replicates), lung cancer (n = 41 tumors, biological replicates), breast cancer (n = 58 tumors, biological replicates), as well as B-lymphoid malignancies, including mantle cell lymphoma (n = 21 tumors, biological replicates), follicular lymphoma (n = 54 tumors, biological replicates), DLBCL (n = 47 tumors, biological replicates) and Hodgkin’s lymphoma (n = 20 tumors, biological replicates). Representative images are shown.

Extended Data Fig. 4. GSK3B-inhibition impairs β-catenin protein degradation.

(a) Human B-ALL (PDX2) and colon cancer (LOVO) cells were treated with LY2090314 for 0-8 hours and Western blot was performed to measure total β-catenin, as well as active, non-phosphorylated β-catenin using β-actin as loading control (n = 3 independent experiments). (b-c) FACS plots are shown for dual reporter cells that were treated with the GSK3B-inhibitor LY2090314 (10 nM; red) or vehicle control (green) for 16 hours. Inhibition of β-catenin protein degradation by LY2090314 was compared in (b) human colon cancer (SW480, HT-29) cell lines and in (c) human B-ALL (NALM6, KOPN8) cell lines by measuring the changes in GFP- compared to mScarlet-signal (2 independent repeats). (d) β-catenin and Myc levels in B-ALL cells (PDX2, MXP2) following 0-6 h WNT3A (400 ng/ml) treatment in comparison to levels achieved following GSK3β inhibition (LY2090314, 20 nM) and colon cancer (SW620) (n = 2 independent experiments). (e) B-ALL xenografts (MXP2, PDX2) were stimulated with WNT3A conditioned media for indicated time points and activation of Wnt/β-catenin pathway was assessed by measurement of luminescence signal by TOP-Flash reporter. Data is representative of 2 independent experiments and shown as mean±s.d (n = 3 technical replicates). (f) Effect of WNT3a (grey circle) stimulation on growth of B-ALL cells (PDX2) that are wild type (CTNNB1 + /+, green line) and knock-out (CTNNB1-/-, red line) for β-catenin compared to vehicle control (white circle). Data shows mean values from 2 independent experiments each with n = 3 technical replicates. Source data

Extended Data Fig. 5. Genetic attenuation of β-catenin protein degradation induces acute cell death in pre-B cells in vitro and in vivo.

(a) Pre-B cells from Gsk3b^fl/+ mice were transduced with 4-OHT-inducible Cre-ER^T2 or ER^T2 along with GFP. Western blot analysis was performed to visualize Gsk3b deletion and β-catenin accumulation in GFP⁺ cells 2 days after 4-OHT mediated deletion of one allele of Gsk3b. Effects of Gsk3b haploinsufficiency and β-catenin accumulation were assessed in competitive growth assays monitoring changes in percentages of GFP⁺ cells for 0-8 days after 4-OHT treatment. Mean values from 2 independent experiments each with n = 2 technical replicates are shown. (b) Apc^fl/+ pre-B cells were transduced with GFP-tagged Cre-ER^T2 or ER^T2 constructs, and Western blot analysis was performed in GFP⁺ cells to confirm Apc levels and accumulation of β-catenin 2 days after 4-OHT treatment. Following loss of one allele of APC, competitive fitness of pre-B cells was assayed by measuring changes in percentages of GFP⁺ cells. Data presents mean values from 2 independent experiments with each n = 3 technical replicates. (c) Ck1a^fl/+ pre-B cells were transduced with Cre-ER^T2 or ER^T2 and GFP. Western blot analysis was performed for Ck1α, β-catenin and β-actin 2 days after 4-OHT mediated CK1α deletion. Growth kinetics of cells with monoallelic deletion of Ck1α (Cre-ER^T2) was compared to cells with wild type Ck1α (ER^T2) by flow cytometry of changes in GFP⁺ cells 0-8 days post 4-OHT treatment. Data shows mean values from 2 independent experiments each with n = 2 technical replicates. (d) Pre-B cells from β-catenin^{(S33-S45)+/fl} mice expressing Cre-ER^T2 or ER^T2 along with GFP were treated with 4-OHT to induce Cre-mediated excision of GSK3B-phosphorylation sites. Accumulation of β-catenin was confirmed by Western blot. Removal of exon-3 results in a shorter form of β-catenin. Competitive fitness of pre-B cells was assessed by measuring the frequencies of GFP⁺ cells by FACS. Data are presented as the mean±s.d of 3 independent experiments each with n = 3 technical replicates. (e-h) β-catenin^{(S33-S45)+/fl} mice were crossed with Mb1^Cre/+ mice for B-cell-specific excision of GSK3B-phosphorylation sites and B-cell development in the bone marrow and peripheral lymphoid organs was studied by FACS. (e) Absolute numbers (f) percentages of B-cell precursor subsets in the bone marrow of the mice are shown for Mb1^Cre/+ β-catenin^{(S33-S45)+/fl} (red, n = 6 mice, biological replicates) and Mb1^Cre/+ β-catenin^(S33-S45)+/+ mice (green, n = 6 mice, biological replicates). Bone marrow B-cell precursors were distinguished as pro-B cells (CD43⁺ B220^low IgM⁻ BP1⁻), pre-BI cells (CD43⁺ B220^low IgM⁻ BP1⁺), pre-BII cells (CD43⁻ B220^low, IgM⁻), immature B cells (CD43⁻ B220^low IgM⁺) and mature B cells (CD43⁻ B220^high IgM⁺). (g) Representative FACS plots and absolute numbers and frequencies of B220⁺ B-cells in spleens of Mb1^Cre/+ β-catenin^{(S33-S45)+/fl} (red, n = 5 mice, biological replicates) and Mb1^Cre/+ β-catenin^(S33-S45)+/+ (green, n = 5 mice, biological replicates) mice are shown. (h) Representative FACS plots, absolute numbers, and fractions [%] of B-cells in the peripheral lymph nodes of Mb1^Cre/+ β-catenin^{(S33-S45)+/fl} (red, n = 4 mice, biological replicates) and Mb1^Cre/+ β-catenin^(S33-S45)+/+ (green, n = 4 mice, biological replicates) mice are shown. Throughout (e-h) two tailed unpaired t-tests were used to calculate P values and data are presented as mean±s.d. Source data

Extended Data Fig. 6. β-catenin forms repressive complexes with Ikaros and NuRD factors in B-cells.

(a-c) Human B-ALL (MXP2), B-cell lymphoma (JEKO), AML (MOLM13), colon cancer (SW480) and lung cancer (H446) cell lines were engineered with β-catenin mutations impairing GSK3B and CK1a-mediated phosphorylation. (a) Proteins bound to β-catenin were identified by mass-spectrometry following Co-IP with antibodies against β-catenin or control Ig. β-catenin binding proteins in each cell type were plotted as a function of background binding (x-axis, non-specific binding defined by CRAPOME database) and log₂-fold enrichment over control Ig (y-axis). (b) Western blot was performed to validate β-catenin-binding of Ikaros factors IKZF1, IKZF2, IKZF3 and NuRD complex component MTA2, as well as AXIN1 (positive control) using input lysates and elutes from the Co-IP experiments with antibodies against β-catenin or control Ig (n = 2 biological replicates). (c) Principal component analysis corroborated B-lymphoid-specific (B-ALL, B-cell lymphoma) β-catenin-interactomes that were clustered together along PC1 axis and separated from myeloid, colon and lung epithelial cells. Source data

Extended Data Fig. 7. β-catenin and Ikaros factors mediate recruitment of NuRD complex at BENC enhancer.

(a) ChIP-seq analysis for β-catenin, Ikaros factors Ikzf1 and Ikzf3, histone marks H3K27ac and H3K4me3 shown for the Myc locus, including upstream Myc promoter regions and long-range transcriptional enhancers of Myc in B-ALL cells from β-catenin^{(S33-S45)+/fl} mice. Peak density plots show colocalization of β-catenin, Ikzf1 and Ikzf3 peaks and their concentration at BENC enhancer regions. Comparison of changes in H3K27ac active enhancer and H3K4me3 active promoter marks following accumulation of β-catenin (red boxes) or β-catenin baseline (light green boxes), upon Ikaros factor deletion (empty boxes) or Ikaros baseline (dark green boxes) conditions. Ikaros factors and β-catenin show marked enrichment at BENC regions. While accumulation of β-catenin depleted H3K27ac marks at BENC enhancer regions, β-catenin had the opposite effect and increased BENC enhancer activity and H3K27ac signals when Ikaros factors (Ikzf1 and Ikzf3) were deleted. (b) β-catenin binding (mean value of two technical replicates) to BENC enhancer region and other β-catenin targets was measured by ChIP-seq in β-catenin^{(S33-S45)+/fl} B-ALL cells (Ikzf1/3 WT) 1 day after β-catenin stabilization. β-catenin is strongly enriched at BENC enhancer (input n = 11863, Other n = 11857, BENC n = 6 peaks; two-sided unpaired t-test P = 0.0005). Boxplots display boxes at first and third quartile with line at median and whiskers from minimum to maximum datapoints within 1.5x interquartile range. (c) ChIP-qPCR was performed for NuRD complex components (MTA2 and CHD4) at the epithelial Myc enhancer regions as well as lymphoid BENC Myc enhancer region (BENC-D). For each region, MTA2 and CHD4 ChIP were performed for B-ALL cells with induced β-catenin accumulation (red boxes), β-catenin deletion (empty boxes) or intact β-catenin (light green boxes), as well as deletion of Ikaros factors (empty boxes) or intact Ikaros factors (dark green boxes). Data shown represents mean values from 3 independent experiments (n = 3 technical replicates). Two-sided unpaired t-test was used to calculate P values. Source data

Extended Data Fig. 8. Ikaros factors compete with TCF family transcription factors for binding to β-catenin.

a) Changes of β-catenin peaks as well as H3K4me3 and H3K27ac histone marks were assessed in the presence (green boxes) and absence of Ikaros factors (empty boxes) and inducible accumulation of β-catenin (red boxes). Heatmaps to illustrate genome-wide distribution of β-catenin peaks, H3K27ac and H3K4me3 histone marks. Ikzf1 and Ikzf3 deletion enabled binding of β-catenin to inactive enhancers and their subsequent activation (gained H3K27ac). Upon loss of Ikzf1 and Ikzf3 regions that have decreased (n = 1,020) or lost (n = 1,675) β-catenin binding are strongly enriched for Ikaros motifs (P = 6.5E-65). Deletion of Ikaros factors resulted in redistribution of β-catenin to canonical Tcf7 (P = 1.0E-42), Tcf7l2 (P = 1.0E-17) and Tcf7l1 (P = 1.0E-16) motifs and depletion from Ikaros motif (P = 6.5E-65), Fisher’s Exact test. b) Changes in β-catenin interactomes in mouse B-ALL cells upon Ikaros factor deletion (gIkzf1/3) were analyzed by co-IP and Western blot analysis in whole cell lysates (input), proteins bound (elute) and flow through (FT) after Co-IP with antibodies against β-catenin or control antibodies. Co-IP was performed under conditions of β-catenin accumulation (red box) and in the presence of Ikaros factor deletion (empty box) or Ikaros baseline levels (dark green box). Deletion of Ikaros factors enabled binding of β-catenin to Tcf7 and Tcf7l2 (n = 2 biological replicates). (c) Scenario of transcription factor complexes with β-catenin in B-lymphoid vs other cell types: TCF factors pair with β-catenin for transcriptional activation of Myc at Wnt responsive elements (WRE) and epithelial enhancer regions (left). In B-cells, Ikaros factors outcompete TCF to bind to β-catenin. Ikaros factors and β-catenin recruit of repressive NuRD complexes to B-lymphoid BENC enhancer regions of Myc, resulting in transcriptional repression of Myc (middle). Loss of Ikaros factors enables interactions between β-catenin and TCF factors to restore transcriptional activation of Myc (right). Source data

Extended Data Fig. 9. Repurposing clinically approved inhibitors targeting β-catenin degradation.

(a) Drug responses (AUC) to the GSK3β-inhibitor (CHIR99021) were plotted for lymphoid (n = 98 cell lines, biological replicates) and solid (n = 298 cell lines, biological replicates) tumors (CTD2 screen). Data are represented as mean±s.d and two-sided unpaired t-test was used to calculate significance. (b) To assess whether LY2090314 induced any changes in epithelial tissues, histopathological evaluation was performed in colon, lung, liver and kidneys of the mice that received vehicle control or LY2090314 (n = 4 biological replicates/group) and representative images of H&E staining are shown. (c) To characterize phenotypic changes induced by LY2090314 treatment, B-ALL cells harvested from LY2090314 (red, n = 9 mice, biological replicates), or vehicle (green, n = 10 mice, biological replicates) treated mice were analyzed by FACS for surface expression of CD20, CD19, CD10, CD34, IL7R, TSLPR, VPREB1 and CD117. Fold-changes in mean fluorescence intensities (MFI [FC], y-axis) following LY2090314 treatment compared vehicle control are shown. No major differences were observed (two-sided unpaired t-test). Data are presented as mean±s.d. Source data