Abstract

   Antigen escape relapse is a major challenge in targeted
   immunotherapies, including CD19- and CD22-directed chimeric antigen
   receptor (CAR) T-cell for B-cell acute lymphoblastic leukemia (B-ALL).
   To identify tumor-intrinsic factors driving antigen loss, we perform
   single-cell analyses on 61 B-ALL patient samples treated with CAR T
   cells. Here we show that low levels of IKAROS in pro-B-like B-ALL cells
   before CAR T treatment correlate with antigen escape. IKAROS^low B-ALL
   cells undergo epigenetic and transcriptional changes that diminish
   B-cell identity, making them resemble progenitor cells. This shift
   leads to reduced CD19 and CD22 surface expression. We demonstrate that
   CD19 and CD22 expression is IKAROS dose-dependent and reversible.
   Furthermore, IKAROS^low cells exhibit higher resistance to CD19- and
   CD22-targeted therapies. These findings establish a role for IKAROS as
   a regulator of antigens targeted by widely used immunotherapies and in
   the risk of antigen escape relapse, identifying it as a potential
   prognostic target.

   Subject terms: Haematological cancer, Tumour immunology
     __________________________________________________________________

   Antigen escape relapse remains a major obstacle when treating B-cell
   malignancies with immunotherapies. Here, the authors show that IKAROS
   regulates the surface expression of CD19 and CD22, revealing a
   mechanism by which B-cell cancers evade targeted immunotherapies.

Introduction

   B-cell acute lymphoblastic leukemia (B-ALL) is the most common
   childhood cancer, accounting for 25% of cancer diagnoses in children
   and adolescents up to 19 years of age^[74]1. Approximately 15% of
   children and young adults diagnosed with ALL will relapse, and 50% of
   children who relapse will die, making relapsed B-ALL the second leading
   cause of cancer-related death for children in the U.S^[75]1. Adult
   B-ALL has a dismal prognosis, with only 40% of patients alive 5 years
   from diagnosis^[76]2. CD19-directed CAR T (CART19) cells are now
   standard therapy for children and adults with relapsed or refractory
   B-ALL, achieving remission rates of 70 – 90%^[77]3,[78]4.
   Unfortunately, around 50% of initial responders will eventually
   relapse, many with CD19 loss^[79]5,[80]6. Similarly, CD19^neg relapses
   have been reported in patients treated with blinatumomab, an
   anti-CD3/CD19 bispecific T cell engager^[81]7,[82]8. For patients
   suffering CD19^neg relapse, CD22-targeted therapies, either inotuzumab
   ozogomycin or CD22-directed CAR T (CART22) cells, have emerged as a
   salvage option^[83]9. However, CD22 downregulation has limited durable
   remissions^[84]10. Several mechanisms of CD19 loss have been reported,
   including truncated CD19 mutations^[85]11, disruption of CD19
   trafficking to the cell surface^[86]12,[87]13, CD19 mRNA
   mis-splicing^[88]14,[89]15, and lineage switch^[90]16; fewer studies
   have focused on the mechanisms behind CD22 downregulation, but
   alternative splicing has also been reported^[91]17. In several studies,
   CD19 loss was accompanied by CD22 downregulation^[92]18–[93]22.
   Recovery of CD19 and CD22 expression has been reported after relief of
   immune pressure^[94]17,[95]18,[96]23,[97]24.

   To discover cell-intrinsic factors associated with antigen loss, we
   analyzed 39 pre- and post-CART19 patient-derived xenografts (PDXs) from
   25 patients using mass cytometry (CyTOF) and single-cell RNA and
   antibody tag sequencing (CITE-seq). In pre-CART19 samples, we
   identified CD19^pos pro-B-like tumor cells with low expression of the
   transcription factor IKAROS to be associated with CD19^neg relapse. We
   determined that IKAROS regulates CD19 expression in B-ALL, large B cell
   lymphoma (LBCL), and chronic lymphoblastic lymphoma (CLL) models.
   IKAROS^low cells demonstrate a wholesale change in chromatin and
   transcriptional state, shifting their identity away from B cells and
   moving towards myeloid and progenitor states. This loss of B cell
   identity manifests with decreased CD19 and CD22 surface expression.
   Consistent with this, we confirmed low IKAROS levels are also
   associated with CD22^low relapse in an independent cohort of 11 r/r
   B-ALL patients treated with CART22. IKAROS-mediated decrease in CD19
   and CD22 surface expression confers a survival advantage for IKAROS^low
   B-ALL cells against CD19- and CD22-targeted therapies. Together, we
   describe a role for IKAROS in immunotherapy target regulation and
   implicate pre-existing IKAROS^low cells at risk of antigen loss under
   the pressure of CD19- or CD22-targeted therapies in patients with
   B-cell malignancies.

Results

CD19 loss is accompanied by loss of B cell features

   To identify tumor-intrinsic factors associated with CD19^neg relapse
   following CAR T therapy, we profiled 35 patient-derived xenograft (PDX)
   samples from pediatric and adult leukemia patients treated with 19.BBz
   CAR T cells using CyTOF (Fig. [98]1A, Supplementary Tables [99]1 and
   [100]2). This cohort includes pre-CART19 PDX samples from patients who
   achieved durable complete remission (CR; n = 6), underwent CD19^neg
   (n = 11) or CD19^pos (n = 4) relapse, or were non-responders (n = 4).
   We also analyzed paired post-CART19 relapse PDX samples from 14 of
   these patients, including those with CD19^neg relapse (n = 8), CD19^pos
   relapse (n = 2), or non-response (n = 4). Evaluation of CD19 expression
   in pre-CART19 PDX samples did not show differences across the clinical
   groups (Fig. [101]1B).

Fig. 1. CD19 loss is associated with loss of B cell identity.

   [102]Fig. 1
   [103]Open in a new tab

   A CART19 cohort of PDX samples analyzed by CyTOF. Created in
   BioRender^[104]81. B Median CD19 expression in CR (n = 6), Negative
   (n = 11), Positive (n = 4), and Refractory (n = 4) pre-CART19 samples.
   C UMAP based on developmental classifier protein expressions in Lin-/
   B+ fraction of healthy BM (left) and projection of tumor cells from
   pre-CART19 patients that achieved durable CR (n = 6), suffered CD19^neg
   relapse (n = 11), and post-CD19 loss (n = 8), respectively. IBI
   Immature-BI, IBII Immature-BII, IBIII Immature-BIII, MBI Mature-BI,
   MBII Mature-BII, ENBI Early-non-BI, ENBII Early-non-BII, NBIII
   Non-BIII.. D Developmental classification of samples in (C). The
   pro-BII population is significantly more abundant in the CR group (CR
   vs. Negative pre-CART19 p-value: 0.0005; CR vs. Negative post-CART19
   p-value: 0.0017), while early-non-BI is significantly more abundant
   between the pre- and post-CD19^neg relapse groups (p-value = 0.0202). E
   Median protein expression of CD19, IKAROS, and PAX5 in pro-BII-like
   tumor cells in CR (n = 6), Negative pre-CART19 (n = 11), and Negative
   post-CART19 (n = 7) samples. F Cohort of pre-and post-CART19 B-ALL
   samples analyzed by single-cell CITE-seq. G UMAP based on most
   variables genes in Lin-/ B+ fraction of healthy BM. Space occupied by
   healthy clusters is depicted. H Projection of B-ALL cells from 9
   samples onto healthy BM-defined UMAP space. I B-ALL cells from CR
   (n = 3), Negative pre-CART19 (n = 3), and Negative post-CART19 (n = 3)
   samples were associated with their closest healthy cluster based on
   k-nearest neighbor assignment. Cluster 6 is significantly less abundant
   in the Negative pre-CART19 group (Negative pre-CART19 vs. CR
   p-value < 0.0001; Negative pre-CART19 vs. Negative
   post-CART19 < 0.0001), while cluster 9 is significantly more abundant
   in the Negative pre-CART19 group (Negative pre-CART19 vs. CR
   p-value < 0.0001; Negative pre-CART19 vs. Negative
   post-CART19 < 0.0001). J–K IKZF1 gene expression (J) and single-cell
   enrichment score for HSC/MPP gene signature (K) in pro-B-like B-ALL
   cells (cluster 6). N represents individual patient samples. Patient
   samples analyzed by mass cytometry or single-cell CITE-seq were not
   performed in replicates. Boxplots in (B) and (E) extend from the 25th
   to the 75th percentiles, with a line in the middle representing the
   median and whiskers extending from the minimum to the maximum values.
   Curves in (D) and (I) show mean ± SEM. Violin plots in (J) and (K) show
   the median (solid line) and 25^th and 75^th quantile (dash lines). The
   statistical tests used were one-way ANOVA followed by Tukey’s multiple
   comparisons tests (B); two-way ANOVA followed by Tukey’s multiple
   comparisons tests (D), (E), and (I); and two-sided Wilcoxon rank sum
   test followed by Bonferroni’s multiple comparisons test in (J) and (K).

   We applied our B-cell developmental classifier to compare cell types
   across patients^[105]25. Consistent with our prior findings in de novo
   B-ALL^[106]25, most leukemic cells were classified within the pre-pro-B
   to pre-BI transitional populations, particularly as pro-BII-like B-ALL
   cells (Fig. [107]1C–D). Interestingly, after CD19 loss, a significant
   fraction of B-ALL cells were classified as early-non-BI (Fig. [108]1D).
   Early-non-BI cells are progenitor cells that do not express CD19 or
   other pan B-cell markers (Supplementary Fig. [109]1B). Enrichment in
   the early-non-BI population after CART19 was not observed in patients
   with CD19^pos relapse (n = 2) or refractory patients (n = 4;
   Supplementary Fig. [110]2A–B). This enrichment was not observed when
   comparing isogenic CD19 wild-type (WT) or knockout (KO) B-ALL cell
   lines (Supplementary Fig. [111]2C), suggesting that the sole loss of
   CD19 expression does not alter the developmental profile of B-ALL
   cells. Instead, in patients, CD19 loss is accompanied by the loss of
   additional B-cell features.

Low levels of IKAROS in pro-B-like tumor cells are associated with CD19^neg
relapse

   Since pro-BII-like B-ALL cells were the most abundant across all
   patient groups (Fig. [112]1D), we compared protein expression in
   pro-BII-like cells from patients who achieved durable CR or underwent
   CD19^neg relapse. While CD19 and PAX5 expression did not differ between
   pre-treatment groups, pro-BII-like cells from patients who would
   experience CD19^neg relapse had lower expression of the B-lineage
   transcription factor IKAROS (Fig. [113]1E). This difference in IKAROS
   level was specific to pro-BII like cells. It was not observed in the
   bulk leukemia cells or any other population (Supplementary
   Fig. [114]2D).

   To further explore the differences between patients who achieved
   durable CR and those who underwent CD19^neg relapse, we performed
   CITE-seq on cells from healthy BM (n = 1), pre-CART19 samples from
   patients who achieved CR (n = 3), and paired pre- and post-CD19^neg
   relapse samples (n = 3) (Fig. [115]1F, Supplementary Table [116]3). We
   defined 18 cell populations in the healthy BM based on gene and protein
   expression profiles (Fig. [117]1G, Supplementary Fig. [118]3A,
   Supplementary Data [119]1). Then, we projected each B-ALL cell onto the
   healthy BM space (Fig. [120]1H) and assigned its closest healthy
   population. Most B-ALL cells were associated with healthy hematopoietic
   multipotent progenitor cells (HSC/MPP, cluster 9) or pro-B cells
   (cluster 6) (Fig. [121]1I). Although bulk samples and HSC/MPP-like
   B-ALL cells showed no differences in gene expression across clinical
   outcomes (Supplementary Fig. [122]3B–C), pre-CART19 pro-B-like B-ALL
   cells from patients with durable CR or CD19^neg relapse exhibited
   distinct gene signatures (Supplementary Fig. [123]3D). In particular,
   IKZF1 expression was significantly lower in pre-CART19 CD19^neg relapse
   pro-B-like B-ALL cells (Fig. [124]1J, Supplementary Data [125]2). This
   was not true in bulk samples or HSC/MPP-like cells (Supplementary
   Fig. [126]3e). There were no differences in CD19 mRNA, CD19 protein, or
   PAX5 mRNA levels in pre-CART19 pro-B-like B-ALL cells (Supplementary
   Fig. [127]3F), consistent with our findings at the protein level
   (Fig. [128]1E). Pre-CART19 IKZF1^low pro-B-like B-ALL cells from
   patients who suffered CD19^neg relapse were enriched for an HSC/MPP
   gene expression signature, indicating these cells had less B-cell
   identity before CART19 administration (Fig. [129]1K). These features,
   low levels of IKZF1 expression and enrichment in HSC/MPP gene
   expression signature, were conserved after CD19 loss (Fig. [130]1J–K).

IKAROS modulates CD19 splicing and CD19 surface expression

   To investigate the impact of IKAROS on CD19 surface expression, we
   targeted IKAROS in B-ALL cell lines using two separate short hairpin
   RNA (shRNA) sequences (KD1 and KD2, Fig. [131]2A), or lenalidomide, a
   cereblon inhibitor that targets IKAROS for degradation^[132]26
   (Fig. [133]2B), or a combination of both methods (KD1 with or without
   10 µM lenalidomide, Fig. [134]2C). These experiments confirmed that
   decreased IKAROS levels reduced CD19 surface expression. Furthermore,
   we confirmed the relationship between IKAROS and CD19 surface
   expression in LBCL (Supplementary Fig. [135]4A–C) and CLL
   (Supplementary Fig. [136]4D) models, suggesting that IKAROS modulation
   of CD19 surface expression is consistent across B cell malignancies.

Fig. 2. IKAROS regulates CD19 and CD22 surface expression.

   [137]Fig. 2
   [138]Open in a new tab

   A–C Relative IKAROS and CD19 median levels in B-ALL cell lines (697,
   NALM6, NALM16, NALM20, REH, RS4;11, SUP-B15) transduced with lentivirus
   expressing scrambled or short hairpin RNA (shRNA) against IKZF1 (A; n
   from left to right: 20, 20, 8, 25, 25, 12), treated with DMSO or 10 µM
   lenalidomide (B; n = 8), or combining shRNA knockdown with or without
   lenalidomide treatment (C, n = 9). Proteins were measured by flow
   cytometry and normalized to scrambled transduced (A), DMSO-treated (B),
   or scrambled transduced and DMSO-treated (C) cells. RFI = relative
   fluorescence intensity. N represents independent biological replicates.
   D CD19 variance-stabilized transformed (vst) counts in IKAROS WT or KD
   B-ALL cells. E Accessibility profile of CD19 promoter and gene from one
   representative IKAROS WT and KD B-ALL cell line pair. Other cell lines
   and their biological replicates can be found in Supplementary
   Fig. [139]5A. F Gene set enrichment analysis (GSEA) results for RNA
   splicing (GO: 0008380) gene signature in IKAROS WT and KD B-ALL cells.
   G, H Frequency of intron 10 retention in CD19 mRNA in IKAROS WT or KD
   B-ALL cells (G) or pediatric B-ALL patients treated with 19.BBz CAR T
   cells^[140]11,[141]14 (n = 11), adult B-ALL patients treated with
   blinatumomab^[142]13 (n = 2), and adult LBCL patients treated with
   19.28z CAR T cells^[143]29,[144]30 (n = 6) (H). I GSEA results for
   Zheng Cord Blood C6 HSC/MPP gene signature^[145]80 in IKAROS WT and KD
   B-ALL cells. J Cell type enrichment analysis of genes with
   differentially accessible promotors in IKAROS WT or IKAROS KD B-ALL
   cells. K CD22 vst counts in IKAROS WT or KD B-ALL cells. L Relative
   CD22 median levels in isogenic IKAROS WT or KD B-ALL cells (n = 7;
   independent biological replicates). Values were measured by flow
   cytometry or CyTOF and normalized to IKAROS WT condition. M Paired
   pre-CART22 and post-CD22^low relapse cohort of leukemic patient samples
   analyzed by CyTOF. IKAROS median levels in pro-BII-like tumor cells. N
   represents individual patient samples. Schematic representation of the
   patient cohort created in BioRender^[146]81. RNA-seq and ATAC-seq
   experiments (D–G, I–K) were performed in 3 cell lines (NALM6, REH, and
   SUP-B15) with two biological replicates per cell line. Bar plots in
   (A–D), (G, H), and (K, L) show mean ± SEM. Boxplots in (M) extend from
   the 25th to the 75th percentiles, with a line in the middle
   representing the median and whiskers extending from the minimum to the
   maximum values. Statistical tests used were two-way ANOVA followed by
   Šidák’s multiple comparisons test (A–C); DESeq’s Wald test followed by
   BH correction (D) and (K); multivariate analysis of transcript splicing
   (rMATS) followed by FDR correction (G); and two-sided paired t-test (L)
   and (M). Not significant (n.s.), P > 0.05.

   To understand how IKAROS modulates CD19 surface expression, we analyzed
   isogenic IKAROS wild-type (WT) and knockdown (KD) B-ALL cell lines
   using ATAC-seq and RNA-seq (Supplementary Fig. [147]4E). Transcriptomic
   data showed similar CD19 mRNA levels in IKAROS WT and KD cells
   (Fig. [148]2D). Chromatin accessibility at the CD19 promoter and gene
   locus also showed no changes (Fig. [149]2E, Supplementary
   Fig. [150]5A). However, gene set enrichment analysis (GSEA)
   demonstrated enrichment of genes associated with RNA splicing in IKAROS
   WT cells, suggesting potential alterations in splicing events in IKAROS
   KD cells (Fig. [151]2F). CD19 mis-splicing has been reported to
   influence CD19 expression^[152]14,[153]15. In particular, CD19 intron 2
   retention, which encodes a truncated CD19 protein, has been reported in
   healthy BM, pre-CART19 leukemic blasts, and associated with CD19^neg
   relapse^[154]15,[155]27,[156]28. While baseline retention of CD19
   intron 2 was already high in B-ALL cell lines (79.6%), IKAROS KD B-ALL
   cells showed a trend toward higher retention (86.1%; Supplementary
   Fig. [157]6D). Interestingly, a different splicing event, CD19 intron
   10 retention, was significantly higher in IKAROS KD cells
   (Fig. [158]2G). We observed more CD19 intron 10 retention in CD19^neg
   relapsed samples from pediatric B-ALL patients treated with 19.BBz CAR
   T cells^[159]11,[160]14 (n = 11), adult B-ALL patients treated with
   blinatumomab^[161]13 (n = 2), and adult LBCL patients treated with
   19.28z CAR T cells^[162]29,[163]30 (n = 6) (Fig. [164]2H, Supplementary
   Fig. [165]6E). CD19 intron 10 retention was confirmed to occur in 282
   diagnosis or relapse B-ALL patient samples from the TARGET dataset
   (Supplementary Fig. [166]6A). Of note, the annotated CD19 transcript
   retaining intron 10 also retains intron 2 (ENST00000565089,
   Supplementary Fig. [167]6B). Analysis of direct long-read RNA-seq data
   from one B-ALL patient (SRR14326969), confirmed that CD19 intron 10 was
   co-retained with intron 2 (Supplementary Fig. [168]6C). CD19 intron 2
   retention was higher in CD19^neg relapsed pediatric and adult B-ALL,
   and LBCL patients (Supplementary Fig. [169]6E). To assess the impact of
   CD19 intron 10 on CD19 surface expression, we introduced different CD19
   isoforms into CD19 KO B-ALL models (Supplementary Fig. [170]6F). As
   expected, the presence of CD19 intron 2 completely abolished CD19
   surface expression, whereas inclusion of CD19 intron 10 had no effect
   (Supplementary Fig. [171]6G), suggesting that intron 10 retention may
   serve as a proxy for intron 2 retention.

IKAROS is required to sustain B-cell identity and CD22 surface expression

   ATAC-seq analysis identified 1250 and 4037 peaks significantly more
   accessible in IKAROS WT and KD cells, respectively (Supplementary
   Data [172]3), consistent with IKAROS’ role as a transcriptional
   repressor^[173]31. The transcription factor binding sites in these
   differentially accessible peaks differed between IKAROS WT and KD cells
   (Supplementary Fig. [174]7A), indicating altered gene expression
   programs in IKAROS KD cells. Transcriptomic analysis confirmed
   differential expression of several transcription factors between IKAROS
   WT and KD cells (Supplementary Fig. [175]7B, Supplementary
   Data [176]4). IKAROS KD cells showed 168 genes with both more
   accessible promoters and higher expression, forming a network
   characterized by non-B lineage genes (red nodes in Supplementary
   Fig. [177]7C), suggesting a loss of B cell identity. Indeed, while the
   gene expression profile and chromatin landscape of IKAROS WT cells are
   consistent with pro-B cell identity (Fig. [178]2J, Supplementary
   Fig. [179]4G), IKAROS KD cells acquired a profile more characteristic
   of HSC/MPP cells (Fig. [180]2I, J).

   Given the re-wiring of transcriptional networks in IKAROS KD cells, we
   investigated the expression of other B-cell proteins. IKAROS KD cells
   showed reduced CD22 mRNA (Fig. [181]2K) and CD22 surface expression
   (Fig. [182]2L). IKAROS KD also demonstrated increased expression of
   alternatively spliced CD22 isoforms that retain intron 2 or skip exon
   12 (Supplementary Fig. [183]6H). Clinical trials have targeted CD22 in
   patients who relapse after CD19-targeted therapies^[184]10. We analyzed
   22 paired patient samples collected before CART22 administration or
   after CD22^low relapse to assess the clinical implications of IKAROS
   and CD22 interaction. Lower IKAROS levels were found in pro-BII-like
   B-ALL cells following CD22^low relapse (Fig. [185]2M, Supplementary
   Table [186]1). These results suggest that IKAROS modulates CD19 and
   CD22 surface expression, and patients with IKAROS^low tumor cells might
   be at higher risk for antigen-loss relapse following both CD19- and
   CD22-targeted therapy.

IKAROS regulates CD19 and CD22 surface expression in a dose-dependent and
reversible manner

   To fine-tune IKAROS levels without the toxicities associated with shRNA
   infection and lenalidomide treatment, we generated IKAROS-regulatable
   models by overexpressing a codon-optimized version of IKAROS fused to a
   degron tag^[187]32 and knocked out the endogenous IKZF1 gene. In this
   system, in the absence of asunaprevir, a hepatitis C virus
   nonstructural protein 3 protease inhibitor, cells express high levels
   of IKAROS as the degron tag is removed. However, in the presence of
   asunaprevir, IKAROS is rapidly targeted for degradation. We generated
   seven models (IKAROS-degron KO1 – KO7) where IKAROS levels can be
   titrated using different concentrations of asunaprevir (Fig. [188]3A).
   CD19 and CD22 surface levels decreased in an IKAROS dose-dependent
   manner (Fig. [189]3B, C). We selected IKAROS-degron KO1 and KO2 models
   for further studies because they express the highest and lowest
   baseline CD19 surface expression, respectively. Upon asunaprevir
   withdrawal, IKAROS, CD19, and CD22 levels were restored
   (Fig. [190]3D–F), demonstrating reversible regulation of CD19 and CD22
   surface expression by IKAROS.

Fig. 3. IKAROS regulates CD19 and CD22 surface expression in a dose-dependent
and reversible manner.

   [191]Fig. 3
   [192]Open in a new tab

   A–C Relative IKAROS (A, n = 7), CD19 (B, n = 7), and CD22 (C, n = 2)
   median levels in IKAROS-degron models treated with increasing doses of
   asunaprevir. Values were measured by flow cytometry and normalized to
   DMSO-treated (0 µM) condition. N represents independent clones. Dots
   represent the mean value from two technical replicates. RFI relative
   fluorescence intensity. D–F IKAROS-degron models were treated with
   10 µM asunaprevir for 7 days. Afterward, asunaprevir was withdrawn, and
   cells were either treated with DMSO (ASU/DMSO) or maintained with 10 µM
   asunaprevir (ASU/ASU) for an additional 28 days. As a control,
   IKAROS-degron models were treated with DMSO for 35 days. Relative
   levels of IKAROS (D), CD19 (E), and CD22 (F) were measured at 9, 14,
   21, and 28 days post-asunaprevir withdrawal by flow cytometry and
   normalized to DMSO-treated condition. The experiment was performed in
   duplicate using two independent clones (n = 2). RFI relative
   fluorescence intensity. G IKAROS-degron models were treated with DMSO
   (D0; n = 6; 2 independent clones in 3 independent experiments) or 10 µM
   asunaprevir for 3 (n = 2 independent clones), 7 (n = 4; 2 independent
   clones in 2 independent experiments), and 22 (n = 2 independent clones)
   days. The protein profile was measured by flow cytometry and CyTOF and
   normalized to the D0 condition. H Projection of IKAROS-degron models
   treated with DMSO (left) or 10 µM asunaprevir (right) for 7 days onto
   the UMAP representation of healthy B cell developmental populations
   based on developmental classifier protein expressions. I Developmental
   classification of samples in (G). The frequencies of Pro-BII and Pre-BI
   populations were statistically different (p-value < 0.001) in the D0
   condition against D3, D7, and D22 conditions. J, K Relative IK6 (A,
   n = 3), CD19 (B, n = 3), and CD22 (C, n = 3) median levels in
   IK6-degron models treated with increasing doses of asunaprevir. Values
   were measured by flow cytometry and normalized to DMSO-treated (0 µM)
   condition. N represents independent clones. Dots represent the mean
   value from two technical replicates. RFI relative fluorescence
   intensity.Bar plots in (A–F) and (J–L) show mean ± SEM. Boxplots in (G)
   extend from the 25th to the 75th percentiles, with a line in the middle
   representing the median and whiskers extending from the minimum to the
   maximum values. Curves in (I) show mean ± SEM. Statistical tests used
   were paired one-way ANOVA followed by Dunnett’s multiple comparisons
   test (A–C) and (J–L); and two-way ANOVA followed by Dunnett’s multiple
   comparisons test (D–G) and (I).

   CyTOF analysis of IKAROS-degron models treated with asunaprevir for 3,
   7, or 22 days showed downregulation of CD19 and CD22 surface expression
   along with mature B-cell (CD24, CD38, CD72) and B-cell differentiation
   (PU.1) proteins in IKAROS^low conditions. Conversely, proteins in
   progenitor (CD10, CD43), myeloid (CD58, CD63, RUNX1), and
   antigen-presenting (HLA-DR) cells were upregulated (Fig. [193]3G).
   These phenotypic changes promote a shift of IKAROS^low cells toward
   more progenitor and immature B-cell states (Fig. [194]3H–I), confirming
   previous observations of the transcriptional and chromatin landscapes.

   IKAROS alterations are associated with relapse in de novo
   B-ALL^[195]33. The most common IKAROS alteration is the deletion of the
   four DNA-binding domains, resulting in the dominant negative IK6
   isoform^[196]34. To assess whether patients expressing the IK6
   isoform—who are more likely to experience chemotherapy failure and
   subsequently receive CD19- or CD22-targeted therapies—are at higher
   risk for antigen escape relapse, we generated IK6-degron models
   (Supplementary Fig. [197]8A). Surface CD19 and CD22 molecules were
   similar among IKAROS-degron and IK6-degron models (Supplementary
   Fig. [198]8B). Interestingly, decreasing IK6 expression with
   asunaprevir treatment did not modify CD19 or CD22 surface levels in
   IK6-degron models (Fig. [199]3J–L). CyTOF protein profiling revealed
   that asunaprevir treatment did not alter the expression of CD38, CD58,
   CD72, or RUNX1, while the effects on CD24, PU.1, and HLA-DR expression
   were attenuated (Supplementary Fig. [200]8C). CD45 was the only protein
   that showed a distinct response to asunaprevir treatment, being
   upregulated in IK6-degron cells (Supplementary Fig. [201]8c).
   Similarly, IK6-degron models did not exhibit a developmental shift
   following asunaprevir treatment (Supplementary Fig. [202]8D).
   Consistent with the results of our model, in primary patient data,
   there are no differences in CD19 or CD22 gene or protein expression
   between patients with IKZF1 deletions and those with other B-ALL
   subtypes (Supplementary Fig. [203]8E, F). These results suggest that
   patients with IKZF1 deletions are not more susceptible to CD19 or CD22
   downregulation, and in the context of CD19- and CD22-targeted
   therapies, the response is instead related to wild-type IKAROS dose and
   requires the DNA binding domains.

Low levels of IKAROS confer resistance to CD19- and CD22-targeted therapies

   Since antigen density is crucial for CAR T cell efficacy^[204]35, we
   investigated whether IKAROS^low cells have an advantage against CD19-
   and CD22-targeted therapies. First, we confirmed that low IKAROS levels
   significantly reduced the number of CD19 and CD22 molecules on the
   B-ALL cell surface (Fig. [205]4A–C). Then, we co-cultured asunaprevir-
   or vehicle-treated IKAROS-degron cells with mock, blinatumomab, 19.BBz,
   19.28z, 22.BBz, and dual 19/22.BBz CAR T cells. At three different
   effector-to-target ratios, IKAROS^low cells were more resilient to
   CD19- and CD22-targeted therapies (Fig. [206]4D–H, Supplementary
   Fig. [207]9), with more pronounced resistance at lower
   effector-to-target ratios.

Fig. 4. Cells with low IKAROS expression are resistant to CD19- and
CD22-targeted therapies.

   [208]Fig. 4
   [209]Open in a new tab

   A Median IKAROS expression in IKAROS-degron models (KO1 and KO2)
   treated with DMSO (-ASU, n = 3) or 10 µM asunaprevir (+ASU, n = 3) for
   7 days. Values were measured by flow cytometry and normalized to -ASU
   condition. N represents biological replicates from independent
   experiments. RFI relative fluorescence intensity. B, C CD19 (B) and
   CD22 (C) surface quantitation in IKAROS-degron models (KO1 and KO2)
   treated with DMSO (-ASU, n = 3) or 10 µM asunaprevir (+ASU, n = 3) for
   7 days. N represents biological replicates from independent
   experiments. D–H IKAROS-degron models (KO1 and KO2) were treated with
   DMSO (-ASU) or 10 µM asunaprevir (+ASU) for 7 days. IKAROS-degron cells
   were washed out of DMSO or asunaprevir and co-cultured with
   blinatumomab-treated (D), 19.BBz (E), 19.28z (F), 22.BBz (G), and dual
   19/22.BBz (H) CAR T cells at 1:1, 1:5, and 1:10 E:T ratio. B-ALL cell
   viability was measured at every 2 − 3 h interval via IncuCyte. GFP
   median values were normalized to 0 h condition. RFI relative
   fluorescence intensity. (*) denotes conditions in which both KO1 + ASU
   vs. KO1 -ASU and KO2 + ASU vs. KO2 -ASU comparisons were statically
   significant (p-values < 0.05). Please refer to the source data file for
   the exact p-value at each time point. Experiments in (D–H) were
   performed in triplicate with three different T-cell donors. Dots
   represent the mean value from two (A–C) or three (D–H) technical
   replicates. Bar plots in (A–C) show mean ± SEM. Curves in (D–H) show
   mean ± SEM. Statistical tests used were two-way ANOVA followed by
   Šidák’s multiple comparisons test, with a single pooled variance (A–C);
   and two-way ANOVA followed by Šidák’s multiple comparisons test (D–H).

   To determine if the advantage of IKAROS^low cells against these
   therapies was due to reduced antigen expression and to rule out the
   effect of other IKAROS-regulated genes (Supplementary Fig. [210]4F,
   Supplementary Data [211]4), we developed a model whereby CD19 levels
   were unaffected by IKAROS. We knocked out the endogenous CD19 gene in
   our IKAROS-degron models and reintroduced ectopic expression of
   wild-type CD19. In these models, named CD19KO-FL, asunaprevir treatment
   reduced IKAROS levels without decreasing CD19 surface expression
   (Supplementary Fig. [212]10A), while the changes in other proteins
   persisted (Supplementary Fig. [213]10B). As expected, asunaprevir
   treatment did not provide any advantage against CD19-targeted
   therapies, even at low E:T ratios (Supplementary Fig. [214]10C–E).
   Thus, decreased CD19 and CD22 surface expression due to low IKAROS
   levels reduces the efficacy of CD19- and CD22-targeted therapies and
   B-ALL cell killing, demonstrating that IKAROS^low cells have a survival
   advantage in the face of these therapies (Fig. [215]5).

Fig. 5. Proposed model of IKAROS-mediated antigen escape in the face of CD19-
and CD22-targeted therapies.

   [216]Fig. 5
   [217]Open in a new tab

   Before immunotherapy, IKAROS^low pro-B-like B-ALL cells possess
   chromatin and gene expression states poised for loss of B-cell identity
   while maintaining expression of CD19 and CD22. Under immune pressure,
   IKAROS^high cells maintain their antigen expression, making them more
   susceptible to T cell-mediated killing. Conversely, IKAROS^low cells
   are more likely to downregulate their antigen expression, giving them a
   relative advantage to escape immunotherapies, resulting in antigen
   escape relapse. Created in BioRender^[218]81.

Discussion

   Antigen loss after targeted immunotherapies, including blinatumomab,
   CD19- or CD22-directed CAR T cells, remains a significant clinical
   challenge for patients with r/r B-cell malignancies. There is no
   standard of care for patients experiencing a CD19^neg relapse, and
   outcomes are poor^[219]9,[220]36. Often, CD22 is the next antigen
   targeted, as it is expressed in the majority of
   B-ALL^[221]9,[222]10,[223]37,[224]38. Here, we found that the
   transcription factor IKAROS modulates CD19 and CD22 surface expression
   in a dose-dependent and reversible manner. IKAROS is a transcription
   factor crucial in B cell lineage specification and commitment and is a
   known B-ALL tumor suppressor^[225]39. In B-ALL, genetic alterations
   affecting the IKAROS gene, IKZF1, are associated with poor response to
   front-line therapy and are a factor in relapse risk
   stratification^[226]33. However, a role for IKAROS in the failure of
   CD19- and CD22-targeted therapies has not been described.

   Two non-mutually exclusive models explain the source of CD19^neg tumor
   cells. First, the immune-enrichment model posits that rare pre-existing
   CD19^neg tumor cells are selected while antigen-positive cells are
   eradicated^[227]40,[228]41. A single-center study of 166 pediatric and
   young adult r/r B-ALL cases treated with tisagenlecleucel found that
   the presence of CD19^neg/dim tumor cells prior to therapy did not
   predict nonresponse or recurrence after CART19 therapy^[229]42. Second,
   the immune-pressure adaptive model suggests that some CD19^pos tumor
   cells possess intrinsic properties that favor losing CD19 to CD19
   targeting^[230]13. Our data support the latter model. Using single-cell
   approaches in clinically annotated samples, we found that, prior to
   CART19 therapy, CD19^pos pro-B-like tumor cells with low levels of
   IKAROS were associated with CD19 loss and CD19^neg relapse. Notably,
   IKAROS levels differed only in pro-B-like cells, not in other
   subpopulations, highlighting the power of our single-cell approach.

   IKAROS^low pro-B-like tumor cells are enriched for an HSC/ MPP gene
   signature, demonstrating loss of B-cell commitment. CD19 loss has been
   associated with increased expression of stem/progenitor markers, such
   as CD34 and CD123, suggesting de-differentiation to a developmentally
   earlier state^[231]43. IKAROS KD or regulatable models demonstrated
   that B-ALL cells acquire the HSC/MPP signature through modulation of
   IKAROS. Low levels of IKAROS resulted in the modulation of B-cell
   phenotype with a gain of stem/progenitor and myeloid proteins while
   B-lineage proteins declined. These include targets of immune therapies
   in B-ALL (CD22^[232]44, CD38^[233]45, and CD72^[234]46), suggesting
   more potential for phenotypic plasticity and lineage infidelity.
   Wholesale modulation of B-ALL cell state and phenotype aligns with
   previous observations in cell line models and patients, where CD19 loss
   was accompanied by CD22 downregulation^[235]18–[236]22 and loss of both
   antigens is reversible^[237]17,[238]23.

   Although not addressed in our study, lineage switch relapses occur
   after CD19-directed therapy^[239]16. Several studies suggest that the
   acquisition of myeloid features at relapse originates from CD19^pos
   B-ALL cells through reprogramming or selection of clones with HSPC
   features^[240]47–[241]50. Our data suggest that IKAROS^low cells have
   increased plasticity and potential for lineage infidelity and thus may
   be ripe to support lineage switch. It is likely, however, that only
   those with the right conditions (genomic background, cytokine
   signaling) will be able to overcome the barrier to differentiate
   towards the myeloid compartment, consistent with the model presented by
   Jacoby et al.^[242]49.

   Identification of several molecular mechanisms underlying antigen loss
   has emerged with the usage of targeted therapies. CD19 loss has been
   associated with truncating mutations of the CD19 gene^[243]11,
   mutations in genes involved in CD19 membrane
   trafficking^[244]12,[245]13, and CD19 mRNA alternative
   splicing^[246]14,[247]15. CD22 downregulation, which is less understood
   than CD19 loss, has been reported through alterations in transcription
   and splicing^[248]17. Alternative splicing is a complex and rapid way
   to reversibly fine-tune the expression of specific genes. Further,
   there is no evidence to suggest that these potential mechanisms are
   mutually exclusive, meaning mutations in CD19 and mis-splicing events
   may co-occur. We found that low levels of IKAROS result in increased
   CD19 mRNA with co-retention of introns 2 and 10 and confirmed the
   increased expression of this isoform in primary patient samples from
   CD19^neg relapsed B-ALL and LBCL patients after blinatumomab^[249]13,
   19.BBz^[250]11,[251]14, and 19.28z^[252]29,[253]30 CAR T therapies.
   Asnani et al.^[254]15 reported that retention of CD19 intron 2
   introduces a premature stop codon generating a truncated CD19 protein.
   Indeed, we found that the inclusion of CD19 intron 2, but not intron
   10, abolishes CD19 surface expression. CD19 intron 2 transcripts are
   highly expressed at baseline in primary B-ALL cells and cell lines,
   however CD19 intron 10 transcripts are expressed at lower baseline
   levels such that smaller increments in intron 10 retained transcripts
   can be more easily detected, suggesting a potential prognostic role for
   CD19 intron 10 retention as a proxy for intron 2 retention and
   emergence of CD19 loss. For CD22, we observed a small but significant
   increased retention of intron 2 and decreased retention of exon 12 in
   the CD22 mRNA in IKAROS KD models. The CD22 intron 2 retained isoform
   results in a truncated CD22 protein, similar to previously reported
   CD22 exon 2 skipped isoforms^[255]17. CD22 exon 12 skipped isoform has
   been associated with more aggressive leukemias through the loss of CD22
   immunoreceptor tyrosine-based inhibitory motifs^[256]51,[257]52. Our
   data suggests the use of alternative splicing as a rapid way to
   modulate phenotypes during lineage decisions that support plasticity in
   leukemia, which requires further study.

   In addition to differences in CD19 and CD22 mRNA splicing in IKAROS KD
   models, we found further distinctions in how IKAROS modulates CD19
   versus CD22 surface expression. While there were no differences in CD19
   mRNA levels or promotor accessibility in IKAROS KD cells, CD22 mRNA
   levels were significantly lower, with changes in the accessibility to
   two peaks in CD22 intronic regions. Moreover, the kinetics for CD22 and
   CD19 surface expression recovery in the regulatable IKAROS model were
   different. These results suggest that IKAROS modulates CD19 and CD22
   surface expression through different mechanisms, which will be the
   subject of future studies. Despite these distinctions, we show that
   IKAROS DNA binding domains are required to modulate CD19 and CD22
   surface expression. IKZF1 alterations occur in 25 to 80% of B-ALL
   cases, with IKZF1 partial or complete deletions as the main
   alterations^[258]34. In our cohort, IKZF1 deletions were only reported
   in two patients, and both were refractory to CART19 (Supplementary
   Table [259]1). The presence of IKZF1 deletions did not correlate with
   lower CD19 or CD22 mRNA or protein levels. Thus, our data suggest that
   patients with IKZF1 deletions are not more prone to antigen escape
   relapse following CD19- or CD22-targeted immunotherapies.

   High disease burden prior to treatment, non-response to blinatumomab,
   and emergence of minimal residual disease are clinical features
   associated with CD19^neg relapse^[260]36,[261]42,[262]53–[263]55. Prior
   blinatumomab induces higher expression of alternative CD19 isoforms,
   decreases CD19 surface expression, and selects tumor cells more fit to
   survive CD19-directed immunotherapies^[264]13,[265]42,[266]55. In our
   CART19 cohort, 5 patients received prior CD19-targeted therapy
   (blinatumomab or CART19); 4 of them suffered CD19^neg relapse, while
   the other achieved a durable CR. High disease burden may increase the
   frequency of IKAROS^low tumor cells. Furthermore, when challenged with
   CD19- and CD22-targeted therapies, IKAROS^low tumor cells had the most
   survival advantage at lower effector-to-target ratios, conditions that
   resemble higher tumor burdens, and indicate a higher probability for
   IKAROS^low tumor cells to survive and enable antigen escape relapse.

   FDA approval for early administration of blinatumomab in front-line
   therapy and inotuzumab after the first non-response or relapse
   increases the population at risk for antigen escape. Hematopoietic stem
   cell transplantation (HSCT) has been used as a consolidative therapy to
   avoid antigen escape relapse but still risks long-term complications
   and side effects^[267]56. Understanding the mechanisms of antigen loss
   and thereby identifying patients at risk for antigen escape relapse
   would change the treatment paradigm for these challenging patients. Our
   results provide insight into mechanisms of antigen modulation and
   lineage plasticity mediated by IKAROS dose and demonstrate a potential
   prognostic role for IKAROS in the early identification of patients at
   risk of antigen loss and identify IKAROS as a target to revert or
   subvert these relapses.

Methods

Bone marrow samples from patients and healthy donors

   Healthy human bone marrow (BM) was purchased from AllCells, Alameda,
   CA, USA (n = 6; median age was 23 years (range, 21 – 32 years); 5 males
   and 1 female). De-identified patient-derived xenograft (PDX) samples
   were obtained from Children’s Hospital of Philadelphia. thirty-nine
   samples were collected under informed consent from twenty-three
   patients enrolled in CHP959 ([268]NCT01626495^[269]57; n = 18), 14BT022
   ([270]NCT02228096^[271]58; n = 4), and 16CT022
   ([272]NCT02906371^[273]59; n = 1) studies; and two adult patients
   treated compassionately: one patient with r/r B-ALL that underwent
   CD19^pos relapse after CART19 treatment (Patient ID: Pos1) and one
   patient with chronic lymphocytic leukemia/small lymphocytic lymphoma
   (CLL/SLL) with Richter’s transformation that underwent CD19^neg relapse
   after CART19 therapy (Patient ID: Neg11). All patients in this cohort
   were treated with 19.BBz CAR T cells. Twenty-two de-identified primary
   patient samples were obtained from 11 patients who were consented and
   treated with 22.BBz CAR T cells on a phase I trial
   ([274]NCT02315612^[275]60) at the National Cancer Institute. The
   Stanford University Institutional Review Boards approved the use of
   these samples. Clinical data for these samples are available in
   Supplementary Table [276]1.

Cell culture and CD19 knock out generation

   NALM6 (CRL-3273), REH (CRL-8286), RS4;11 (CRL-1873), and SUP-B15
   (CRL-1929) cell lines were purchased from ATCC (Manassas, VI, USA). 697
   (ACC 42), CI (ACC 770), JVM-2 (ACC 12), MHH-CALL4 (ACC 337), NALM16
   (ACC 680), NALM20 (ACC 681), and WA-OSEL (ACC 767) were purchased from
   DSMZ (Braunschweig, Germany). OCI-Ly1, OCI-Ly-7, and SUDHL6 were a gift
   from the Amengual lab^[277]61. 697, JVM-2, NALM6, NALM16, REH, RS4;11,
   SUDHL6, and WA-OSEL were cultured in RPMI-1640 medium supplemented with
   10% fetal bovine serum (FBS). OCI-Ly1 and OCI-Ly-7 were cultured in
   IMDM medium supplemented with 10% FBS. CI, MHH-CALL4, NALM20, and
   SUP-B15 were cultured in IMDM or RPMI-1640 medium supplemented with 20%
   FBS. For all cell lines, the medium was additionally supplemented with
   2 mM L-glutamine (Invitrogen) and 1x penicillin/streptomycin
   (Invitrogen), and cells were maintained at 37 °C and 5% CO[2].

   To generate CD19 KO cell lines, we incubated 3.2 µg synthetic gRNA
   against the CD19 gene (Synthego KO kit v2) with 6 µg purified Alt-R
   S.p. HiFi Cas9 Nuclease V3 (IDT) for 20 min at 37°C. Guide sequences
   were as follows: 5′-UGCCAGGCCUUCUCAGAGGG-3′,
   5′-UUUUAAGAAGGGUUUAAGCG-3′, and 5′-CUUCAACGUCUCUCAACAGA-3′. This
   ribonucleoprotein complex was transfected into 1.5 × 10^5 MHH-CALL4,
   NALM6, REH, or SUP-B15 cells via electroporation using a
   4D-Nucleofector system and AMAXA SF reagent with program CV-104
   (Lonza). Seven days post-electroporation, cells were processed for
   CyTOF and CD19^neg (CD19 KO) cells were gated from CD19^pos (CD19 WT)
   cells.

Mass cytometry

   Samples were processed as previously described^[278]62. For viable
   frozen healthy BM, leukemic primary, and PDX samples, cells were thawed
   in 90% RPMI medium (ThermoFisher Scientific) with 10% FCS, 20 U/ml
   sodium heparin (Sigma-Aldrich), 0.025 U/ml Benzonase (Sigma-Aldrich),
   1x L-glutamine and 1x penicillin–streptomycin (Invitrogen) and rested
   at 37 °C for 30 min. For B-ALL cell lines, cells were resuspended in
   their corresponding culture media. For healthy BM, leukemic primary,
   and PDX samples, 1 × 10^6 cells were stained for viability with
   cisplatin as described^[279]63. Following viability staining, cells
   were fixed with 1.6% paraformaldehyde (PFA, Electron Microscopy
   Sciences) for 10 min at RT. Cells were barcoded using 20-plex palladium
   barcoding plates prepared in-house as described^[280]64. We included
   one healthy BM reference sample within each barcoding plate to control
   for batch effects. A total of 15 barcode plates were used in this
   study. Following barcoding, cells were pelleted and washed once with
   cell-staining medium (CSM; PBS with 0.5% BSA and 0.02% sodium azide) to
   remove residual PFA. Blocking was performed with Human TruStain FcX
   (BioLegend), following the manufacturer’s instructions. Antibodies to
   surface proteins were added, yielding 800-μl final reaction volumes,
   and samples were incubated at RT and 300 g for 30 min (Supplementary
   Table [281]2). Cells were washed with CSM before permeabilization with
   methanol for 10 min at 4 °C. Cells were washed with CSM and stained
   with intracellular protein and phospho-specific antibodies in 800 μl
   for 30 min at RT and 300 g (Supplementary Table [282]2). Cells were
   washed once in CSM, then stained overnight with 1:5,000 191Ir/193Ir or
   103Rh DNA intercalator (Standard Biotools) in PBS with 1.6% PFA at
   4 °C. Cells were washed once with CSM, washed twice with double
   distilled water, filtered to remove aggregates, and resuspended in
   139La/ 142Pr/ 159Tb/ 169Tm/ 175Lu normalization beads^[283]65
   immediately before analysis using a Helios mass cytometer (Standard
   Biotools). Throughout the analysis, cells were maintained at 4 ° C and
   introduced at a constant rate of 150–200 cells/s.

Processing of mass cytometry data

   Data were normalized together using bead normalization^[284]65 and
   files were debarcoded as described^[285]64. Single-cell protein
   expression data were extracted and analyzed using packages from the
   Comprehensive R Archive Network (CRAN) project
   ([286]https://cran.r-project.org/) and Bioconductor
   ([287]http://www.bioconductor.org). Raw data were transformed using the
   inverse hyperbolic sine (arcsinh) function with a cofactor of 5. The
   expression of proteins in each population of interest was determined by
   calculating the median level of expression after arcsinh
   transformation. For UMAP visualization, each population or sample was
   randomly subsampled to 1000 cells. Dimensionality reduction was
   performed using umap package (version 0.2.9.0) based on arcsinh values
   of selected markers.

Manual gating

   Single cells were gated using Community Cytobank software
   ([288]https://community.cytobank.org/) based on event length and
   191Ir/193Ir or 103Rh DNA content (to avoid debris and doublets) as
   described^[289]66. For B-ALL cell lines, live B-ALL cells were gated
   based on cleaved poly(ADP-ribose) polymerase (cPARP) and ^195Pt or
   cleaved Caspase3 (cCaspase3) content^[290]63. Downstream analyses were
   performed in this live B-ALL cell fraction. For healthy BM, leukemic
   primary, and PDX samples, following single-cell gating, live
   non-apoptotic cells were gated based on cleaved poly(ADP-ribose)
   polymerase (cPARP) and ^195Pt content^[291]63. In PDX samples, murine
   cells were excluded by gating on mouse CD45 (mCD45) protein. Platelets
   and erythrocytes were excluded by gating on CD61 and CD235a/b,
   respectively. The remaining fraction was gated to exclude T cells and
   myeloid cells based on CD3e, CD11b, CD16, and CD33 expression. After
   further exclusion of CD38^high plasma cells, the remaining fraction was
   defined as lineage-negative blasts (Lin − / B + ; see Supplementary
   Fig. [292]2A for gating). Further analyses were applied to Lin − / B+
   fraction. For the primary samples in the CART22 cohort, we applied
   opt-SNE^[293]67 to the Lin-/B+ fraction of each paired sample together
   with their corresponding healthy BM reference to gate blast cells, as
   was previously shown^[294]68. The following markers were used for
   opt-SNE: CD10, CD11b, CD16, CD179b, CD24, CD33, CD34, CD38, CD45, CD81,
   HLADR, intracellular immunoglobulin heavy chain (IgMi), surface
   immunoglobulin heavy chain (IgMs), Ki67, phosphorylated CREB,
   phosphorylated S6, and terminal deoxynucleotidyl transferase (TdT).

B-cell developmental classification

   We used the single-cell developmental classifier previously
   reported^[295]25. Briefly, Lin − / B+ fraction from healthy human BM
   was gated into 15 developmental populations of normal B lymphopoiesis,
   mixed progenitors, and mature non-B cell fractions, as shown in
   Supplementary Fig. [296]1. The distribution of each population was
   based on the expression of 10 B cell developmental proteins that were
   used for manual gating: CD19, CD20, CD24, CD34, CD38, CD45, CD127,
   CD179b, IgMi, and TdT. Lin − / B+ cells from each leukemia sample or
   live B-ALL cells from cell lines samples were assigned to the most
   similar healthy fraction based on the shortest Mahalanobis distance
   among distances to all healthy developmental populations in these ten
   dimensions. A cell was designated ‘unclassified’ if none of the
   distances were below the classification threshold (Mahalanobis distance
   = 10, based on the number of dimensions).

Single cell RNA and antibody tag sequencing

   Viably frozen healthy BM and PDX samples (Patient ID: CR2, CR4, CR6,
   Neg2, Neg5, Neg11) were thawed in 90% RPMI medium (ThermoFisher
   Scientific) with 10% FCS, 20 U/ml sodium heparin (Sigma-Aldrich), 0.025
   U/ml Benzonase (Sigma-Aldrich), 1x L-glutamine and 1x
   penicillin–streptomycin (Invitrogen) and rested at 37 °C for 30 min.
   Then cells were filtered through cell strainer size 35 µm and
   centrifuged at 350 g for 5 min. Cells were resuspended in 1 ml of Stain
   Buffer (BD Biosciences) and blocking was performed with Human Fc Block
   (BD Biosciences) following the manufacturer’s instructions. To enrich
   for Lin − / B+ fraction, samples were incubated with biotin-conjugated
   antibodies (Supplementary Table [297]3) for 30 min. Cells were washed
   with Stain Buffer and then incubated with BD Streptavidin Particles
   Plus (BD Biosciences) at the manufacturer’s recommended concentration
   for 30 min at RT. Particle-labeled cells were placed in a magnetic
   holder for 6 min. The supernatant was transferred to a new tube and
   placed back in the magnetic holder for an additional round of depletion
   and supernatant recovery. Cells from the supernatant were then pelleted
   by centrifugation at 350 g for 5 min.

   Lin-/ B+ fraction was resuspended in 180 µl of Stain Buffer
   supplemented with a mix of BD AbSeq oligo conjugated antibodies against
   CD19, CD20, CD24, CD34, CD38, CD45, CD127, and IgM (BD Biosciences,
   Supplementary Table [298]3). Then, each sample was labelled with the
   Human Single Cell Sample Multiplexing kit (BD Biosciences) and
   incubated at RT for 30 min. Cells were washed twice with Stain Buffer
   and resuspended in Sample Buffer (BD Biosciences). Cell number was
   counted with Countess II Cell Counter (Thermo Fisher Scientific). A
   total of 50,000 cells (12,500 cells per sample, up to 4 samples) were
   pooled together, and single cells were isolated in a BD Rhapsody
   cartridge using BD Rhapsody Express Single-Cell Analysis System (BD
   Biosciences). We included one healthy BM reference sample within each
   cartridge to control for batch effects. A total of 3 cartridges were
   used in this study. For each cartridge, we followed the manufacturer’s
   instructions to prepare whole transcriptomic, antibody tag, and sample
   tag libraries using the Whole Transcriptome Analysis (WTA)
   Amplification Kit (BD Biosciences). Libraries from the same cartridge
   were indexed with identical Illumina sequencing adapters. Final
   libraries were pooled together, sequencing on a NovaSeq 6000 sequencer
   (Illumina) at MedGenome (Foster City, CA, USA) with paired-end 100 base
   pair (bp) reads.

Processing of CITE-seq data

   Fastq files were processed through the Rhapsody analysis pipeline (BD
   Biosciences) on the Seven Bridges platform
   ([299]https://www.sevenbridges.com), following the manufacturer’s
   recommendations. Reads were mapped to the hg38 reference genome using
   bowtie2. Final expression matrices contain recursive substation error
   correction (RSEC) adjusted molecule counts per cell in a CSV format.
   Molecule count tables were read into the R package Seurat (version
   3.2). Out of 35,716 total cells, 2106 (5.9%) were multiplets, and 637
   (1.8%) were undetermined events. The mean number of cells sequenced per
   sample was 2748. The mean number of genes detected per cell was 947,
   with a mean of 2670 reads/cell for the mRNA library and 1493 reads/cell
   for the antibody tag library. Cells with less than 500 expressed genes
   or 200 gene reads assigned and more than 50% of genes being
   mitochondrial genes were excluded for downstream analysis, removing
   12,597 cells. We normalized transcriptomic data through variance
   stabilizing transformation with the removal of mitochondrial gene
   percentage as a potential confounding source of variation. Antibody tag
   data was normalized with a centered log-ratio transformation.

   We performed the Wilcoxon rank sum test for differential expression
   analysis followed by Bonferroni correction. Genes and antibody tags
   expressed in at least 10% of cells from one of the conditions under
   comparison, had an absolute fold change higher or equal to 25%, and had
   adjusted p-value lower than 0.05 were called significant. Heatmaps were
   plotted using the ComplexHeatmap package (version 2.6.2), and the mean
   expression of each gene or antibody tag per population. To estimate the
   single-cell enrichment score for the HSC multipotent progenitor
   program, we used the function AddModuleScore from the Seurat package.

Projection of tumor cells onto healthy BM UMAP space

   To define healthy BM populations and their UMAP space, we performed
   principal component (PC) analysis on 3000 most variable genes (MVG)
   across all healthy BM cells and antibody tag data. Based on the top 30
   PCs, we performed UMAP embedding of healthy cells, as well as
   unsupervised clustering with the functions FindNeighbors and
   FindClusters from the Seurat package. We used the function FindMarkers
   to identify genes significantly up-regulated in each cluster against
   all the other clusters. Differentially expressed genes were analyzed
   with Enrichr^[300]69 ([301]https://maayanlab.cloud/Enrichr/) to define
   biological pathways associated with each cluster and assign their
   corresponding cell population.

   To project tumor cells onto the healthy BM space, we predicted their
   top 30 PCs based on the PC analysis performed for the healthy BMs and
   the expression of 3000 MVG defined previously and antibody tag data.
   These predicted top 30 PCs were used to project and embed tumor cells
   onto healthy BM UMAP space. To associate each tumor cell with their
   closest healthy population, we used a k-nearest neighbors model (k-NN).
   Briefly, we performed 10-fold cross-validation to train a k-NN model to
   predict the cell population of healthy BM cells based on the top 30
   PCs. For this model, the ten nearest neighbors were considered, and the
   decision was taken based on majority voting. Finally, we defined the
   closest population to each tumor cell by applying this k-NN model to
   their predicted top 30 PCs.

Retrovirus and Lentivirus production

   For retrovirus production, 293GP cells were used (graciously provided
   by Dr Garry Nolan). Briefly, 70% confluent 293GP on 10-cm poly-d-lysine
   coated plates were cotransfected with 9 μg of
   IKAROS-degron-P2A-mNeonGreen, IK6-degron-P2A-mNeonGreen,
   CD19WT-IRES-eGFP, CD19Int2-IRES-eGFP, CD19Int10-IRES-eGFP,
   CD19Int2-10-IRES-eGFP, 19.BBz, 19.28z, 22.BBz, or 19/22.BBz CAR
   encoding vectors and 4.5 μg RD114 envelope plasmid (graciously provided
   by Dr. Crystal Mackall) with lipofectamine 3000 (Invitrogen). Viral
   supernatants were collected 48 and 72 h post-transfection, centrifuged
   to separate cell debris from the viral supernatant, and frozen at
   −80 °C for future use.

   For third-generation, self-inactivating lentivirus production, 4 – 5 ×
   10^6 293 T cells (graciously provided by Dr. Crystal Mackall) were
   plated on a 10-cm dish for 24 h before transfection. On the day of
   transfection, a mixture of 1.5 µg pMD2.G envelope plasmid (Addgene
   #12259), 2.25 µg psPAX2 packing plasmid (Addgene #12260), and 3.75 µg
   vector plasmid were cotransfected with TurboFect (Thermo Fisher). For
   the current study, we used the following vector plasmids:
   pLKO-RFP-shCntrl (Addgene #69040), pLKO-RFP-IKZF1-sh2 (Addgene #69041),
   pLKO-RFP-IKZF1-sh3 (Addgene #69042), SMARTvector Inducible
   Non-targeting PGK-TurboRFP ([302]VSC11656, Horizon Discovery
   Biosciences), and SMARTvector Inducible Human IKZF1 PGK-TurboRFP shRNA
   1 (V3SH11252-224727699, Horizon Discovery Biosciences). 48 and 72 h
   post-transfection, the supernatant was collected and centrifuged to
   separate cell debris from the viral supernatant. PEG-it solution
   (System Biosciences Innovation) was added to the viral supernatant, and
   lentivirus particles were concentrated following the manufacturer’s
   protocol. Concentrated lentivirus aliquots (20 µl) were stored at
   −80 °C.

Lentiviral and retroviral transduction

   For lentivirus transduction, 1 × 10^6 cells were incubated with 20 µl
   of concentrated lentivirus particles and TransDux Max reagents (System
   Biosciences Innovation) following the manufacturer’s instructions. 48 h
   post-transduction, cells were washed and incubated in their
   corresponding complete media. When appropriate, 1 µg/ml puromycin
   (Invivogen) was added to the complete media for antibiotic selection
   and 250 ng/ml doxycycline (Sigma) for expression of
   doxycycline-inducible systems. Transduction efficiency was followed by
   measuring RFP expression in a CytoFLEX instrument (Beckman Coulter).

   For retroviral transduction, non-tissue culture treated 12-well plates
   were coated overnight at 4°C with 25 μg/ml retronectin in PBS (Takara).
   Plates were washed with PBS and blocked with PBS supplemented with 2%
   BSA for 15 min. Thawed retroviral supernatant was added at ~1 ml per
   well and centrifuged at 1,800 g for 1.5 h before adding 0.5 × 10^6
   cells. After 48 h, transduction efficiency was checked by GFP
   expression in CytoFLEX instrument (Beckman Coulter). When required,
   transduced cells were enriched by sorting GFP-positive cells in a BD
   FACSAria II SORP Cell Sorter (BD Biosciences).

Lenalidomide treatment

   For targeted degradation of IKAROS protein, 5 x 10^5 tumor cells were
   seeded in duplicate in a 24-well plate with 2 ml complete media
   supplemented with 0 µM, 0.1 µM, 1 µM, or 10 µM lenalidomide (Selleck
   Chemicals). DMSO was used as a vehicle to adjust the drug volume added.
   After 3, 4, or 7 days, CD19 and IKAROS levels were measured by flow
   cytometry.

Flow cytometry analysis

   To assess CD19 and CD22 surface expression, 0.5 – 1 ×10^5 cells were
   washed with CSM and incubated with 100 µl antibody mix (1 µl APC
   anti-human CD19 (clone: HIB19, BioLegend); or 1 µl APC anti-human CD22
   (clone: HIB22, BioLegend) with or without 1 µl Pacific Blue anti-human
   CD19 (clone: HIB19, BioLegend) in 100 µl CSM) at RT for 10 min. Cells
   were washed twice and resuspended in 200 µl CSM for flow analysis in
   CytoFLEX cytometer (Beckman Coulter).

   To assess IKAROS intracellular levels, 0.5 – 1 × 10^5 cells were washed
   with CSM and fixed with 1.6% PFA (Electron Microscopy Sciences) in CSM
   for 10 min at room temperature (RT). Cells were washed twice with CSM
   and incubated with 1 µl Pacific Blue anti-human CD19 antibody (clone:
   HIB19, BioLegend) in 100 µl CSM at RT for 10 min. Cells were washed
   twice and permeabilized with 100 µl Methanol at 4°C for 10 min. After
   three washes, cells were incubated with 1 µl Alexa Flour 647 anti-human
   IKAROS antibody (clone: 16B5C71, BioLegend) in 100 µl CSM at RT and
   300 g for 30 min. Cells were washed twice and resuspended in 200 µl CSM
   for flow analysis in CytoFLEX cytometer (Beckman Coulter).

   Quantification of CD19 and CD22 surface molecules on cancer lines was
   performed using the BD Quantibrite™ APC Fluorescence Quantitation Kit
   (BD Biosciences).

   Flow cytometry data was analyzed using Community Cytobank software
   ([303]https://community.cytobank.org/). Briefly, forward versus side
   scatter was used to exclude debris, while forward scatter area versus
   width was used for doublet exclusion. When pertinent, tumor cells were
   gated based on GFP (IKAROS-degron, IKAROS-degron CD19KO-FL, IK6-degron,
   and CD19 isoforms overexpressing cells) or RFP (IKAROS WT and KD cells)
   expression. Finally, CD19, CD22, and IKAROS median fluorescent
   intensities (MFI) were calculated in the gated population.

Generation IKAROS KD cells for ATAC-seq and RNA-seq

   5 × 10^6 B-ALL cells (NALM6, REH or SUP-B15) were transduced with
   lentivirus expressing IKZF1 specific (Addgene #69041) or scramble
   (Addgene #69040) shRNA in duplicate. 72 h post-transduction, 1 × 10^5
   cells were used to check transduction efficiency, CD19 and IKAROS
   levels via flow cytometry analysis as described. In all cases,
   transduction efficiencies were over 90%.

   For ATAC-seq, 50,000 cells were processed as previously
   reported^[304]70. For RNA-seq, total RNA was extracted from 3 – 5
   × 10^6 cells using Rneasy kit with Dnase I treatment (QIAGEN),
   following manufacturer instructions. RNA libraries were performed with
   TruSeq stranded mRNA kit (Illumina). ATAC and RNA libraries were
   sequenced on the Illumina NovaSeq 6000 at MedGenome facility with
   paired-end 50 bp or 100 bp reads, respectively.

RNA-seq data processing

   For isogenic IKAROS WT or KD B-ALL cells RNA-seq data, we sequenced an
   average of 42 × 10^6 reads per sample (range: 29 × 10^6 – 60 × 10^6
   reads). Paired-end reads were aligned and quantified using Salmon
   (version 1.2.0) index against the hg38 reference genome. Gencode
   transcript annotations (version 37) were used for the genomic location
   of transcriptomic units. Reads aligning to annotated regions were
   summarized as counts using the R package tximport (version 1.18.0).
   Differential expression analyses between IKAROS WT and KD samples were
   performed using DESeq2 (version 1.30.1)^[305]71. A false discovery rate
   (FDR) cutoff of 0.05 was used for gene selection. Read counts were
   normalized using variance-stabilizing transformation (vst) built into
   the DESeq2 package. Gene set enrichment analysis (GSEA) was performed
   using the GSEA software (Broad Institute).

   Raw fastq files from paired pre- and post-CD19^neg relapse samples from
   pediatric r/r B-ALL patients treated with 19.BBz^[306]14 (n = 2), adult
   r/r B-ALL patients treated with blinatumomab^[307]13 (n = 2), adult r/r
   LBCL patients treated with 19.28z^[308]29,[309]30 (n = 6), and
   diagnosis or relapse B-ALL patients (n = 282) from the Therapeutically
   Applicable Research to Generate Effective Treatments (TARGET)
   initiative were processed as described above.

Splicing event analysis and CD19 intron 10 retention

   We used the Multivariate Analysis of Transcript Splicing^[310]72
   (rMATS; version 4.1.1) to detect and quantify alternative splicing
   events. Raw fastq files were aligned against the hg38 reference genome
   (Gencode transcript version 37) using STAR software^[311]73 (version
   2.7.8a). Bam files were sorted by coordinated and indexed with SAMtools
   (version 1.12). rMATS was run under default configuration with indexed
   bam files as input. Reads spanning exon junctions and reads that do not
   cross an exon boundary were used for splicing quantification. Splicing
   events with FDR values lower than 0.05 were called significant.
   Inclusion level for different splicing events: SE (skipped exon), MXE
   (mutually exclusive exons), A3SS (alternative 3’ splice site), A5SS
   (alternative 5’ splice site), and RI (retained intron) were used to
   perform plots and additional analysis with custom R scripts.

   To re-analyze paired pre- and post-CD19^neg relapse B-ALL samples from
   Orlando et al.^[312]11, since raw fastq files were not available, we
   obtained bam files with only those RNA-seq reads aligned to CD19 genes
   (Accession number: SRP141691). Unfortunately, these bam files had
   undergone Split’N’Trim step, which precludes the use of rMATS. Thus,
   similar to Asnani, et al.^[313]15, we performed a visual inspection of
   CD19-aligned reads in the IGV browser^[314]74 and used a custom script
   to quantify reads aligned to CD19 intron 2 or 10 over reads aligned to
   CD19 exons 2 and 3 or exons 10 and 11, respectively. In addition, out
   of 10 patients with matched data available, we excluded one (patient
   17) from the analysis for the same reason explained in Cortés-López, et
   al.^[315]27.

ATAC-seq data processing

   ATAC-seq data was processed as previously described^[316]75. Briefly,
   adapters were trimmed using cutadapt, and reads were mapped using
   bowtie2 with a max fragment length of 2000 bp to hg38. We then filtered
   for non-mitochondrial reads, mapq > 20, and properly paired reads, and
   removed duplicates using Picard tools. De-duplicated libraries were
   down-sampled to 15 × 10^6 read pairs. Aligned, de-duplicated bam files
   were loaded into R using DsATAC.bam function in the ChrAccR R package.
   Peaks were called using macs2 with the following parameters on Tn5
   insertion sites: -shift -75 –extsize 150 –nomodel –call-summits
   –nolambda -p 0.01 -B –SPMR. The consensus peak set across technical and
   biological replicates was calculated using the getConsensusPeakSet
   function in the ChrAccR R package. The count matrix was calculated as
   insertion counts across samples at consensus peak set regions using the
   ChrAccR regionAggregation function. DESeq2^[317]71 was used to
   calculate differentially accessible peaks. Differentially accessible
   peaks were used to calculate motif enrichment (getMotifEnrichment
   function in the ChrAccR R package) using the human_pwms_v2 motif
   database (from the chromVARmotifs package). Adjusted p value (FDR
   value) was converted to -log(FDR value), and top enriched motifs were
   plotted by -log(FDR value). Genomic localization of consensus peak set
   was identified using the annotatePeak function in the ChIPseekerR
   package and was visualized using the IGV browser^[318]74. Genes
   harboring differentially accessible peaks on their promotor region
   (-3000 to + 3000 bp from transcriptional starting site (TSS)) were used
   for pathways enrichment analysis in Enrichr
   ([319]https://maayanlab.cloud/Enrichr/). We used Venny 2.1
   ([320]https://bioinfogp.cnb.csic.es/tools/venny/) to define
   up-regulated genes that had higher accessibility to their promoter in
   IKAROS KD cells. Protein-protein interaction network between these
   genes was created with STRING^[321]76 ([322]https://string-db.org/),
   and Enrichr ([323]https://maayanlab.cloud/Enrichr/) was used to define
   their lineage associations.

Viral vector construction

   All DNA constructs were visualized using SnapGene software (v.7.2.1;
   Dotmatics), and cloning was performed using In-fusion seamless cloning
   (Takara).

   IKAROS-degron, IK6-degron, CD19WT, CD19Int2, CD19Int10, and CD19Int2-10
   sequences were ordered from IDT. IKAROS-degron and IK6-degron were
   cloned into a XhoI/EcoRI-digested mNeonGreen – firefly luciferase
   vector (graciously provided by Dr. Crystal Mackall)^[324]77. CD19
   isoforms were inserted into SrfI/XhoI digested
   pMSCV-FLAG-hIKAROS-IRES-GFP vector (Addgene 74046).

   19.BBz and 19.28z CAR constructs were previously cloned into the MSGV1
   retroviral vector^[325]77. In brief, both constructs consisted of the
   FMC63 scFv, with 19.BBz comprising the sequence FMC63 scFv - CD8 hinge
   - CD8 transmembrane - 4-1BB costimulatory domain - CD3z and 19.28z
   comprising the sequence FMC63 scFv - CD28 hinge - CD28 transmembrane -
   CD28 costimulatory domain - CD3z. The 22.BBz and 19/22.BBz constructs
   were generated by replacing the FMC63 scFv of the 19.BBz sequence. For
   22.BBz, an scFv derived from the m971 antibody was used. For 19/22.BBz,
   the following sequence was used: GMCSF leader sequence - FMC63 VH - G4S
   linker - m971 VL - Whitlow linker - m971 VH - G4S linker - FMC63 VL.

CD19 isoforms expressing tumor models

   CD19 KO B-ALL (697, NALM6, NALM16, REH) cell lines were generated as
   described above.

   After seven days, CD19^neg B-ALL cells were sorted using a BD FACSAria
   II SORP Cell Sorter (BD Biosciences). Cells were expanded for seven
   more days and then transduced with CD19WT-IRES-eGFP,
   CD19Int2-IRES-eGFP, CD19Int10-IRES-eGFP, or CD19Int2-10-IRES-eGFP
   expressing retrovirus. After three days, transduced cells were sorted
   based on GFP expression using a BD FACSAria II SORP Cell Sorter (BD
   Biosciences). CD19 surface expression was assessed as mentioned above.

IKAROS-degron and IK6-degron tumor models

   5 × 10^5 REH cells were transduced with IKAROS-degron-P2A-mNeonGreen or
   IK6-degron-P2A-mNeonGreen expressing retrovirus. After three days,
   transduced cells were sorted based on GFP expression using a BD
   FACSAria II SORP Cell Sorter (BD Biosciences). Cells were expanded for
   seven days and knocked out endogenous IKZF1gene using the following
   synthetic gRNA guides (Synthego KO kit v2): 5′-UGUCGUAGGGCGUGUCGGAC-3′,
   5′-CAACAACGCCAUCAACUACC-3′, and 5′-ACCACCUCGGAACCGCCCGG-3′, and the
   same protocol as described above. Cells were expanded for seven more
   days before single-cell cloning by limiting dilution into 96-well
   plates. Wells containing cells were grown to dense cultures before
   analysis of endogenous IKZF1 locus knockout efficiency using ICE CRISPR
   Analysis Tool (Synthego). Clones with 100% knockout efficiency were
   further characterized by IKAROS, CD19, and CD22 response to asunaprevir
   treatment.

   To generate IKAROS-degron CD19KO-FL tumor models, endogenous CD19 locus
   was knocked out in the IKAROS-degron models, as described above. After
   seven days, CD19^neg IKAROS-degron cells were sorted using a BD
   FACSAria II SORP Cell Sorter (BD Biosciences). Cells were expanded for
   seven more days and then transduced with CD19WT-IRES-eGFP. After three
   days, transduced cells were sorted based on CD19 expression using a BD
   FACSAria II SORP Cell Sorter (BD Biosciences).

Asunaprevir treatment

   A total of 5 x 10^5 IKAROS-degron, IK6-degron, and IKAROS-degron
   CD19KO-FL tumor cells were seeded in duplicate in a 24-well plate with
   2 ml complete media supplemented with 0 µM, 0.1 µM, 1 µM, or 10 µM
   asunaprevir (Selleck Chemicals). DMSO was used as a vehicle to adjust
   the drug volume added. After 3, 7, or 22 days, CD19, CD22, and IKAROS
   levels were measured by flow cytometry. When indicated, 1 × 10^6 cells
   were processed for CyTOF, as described above.

   For expression recovery experiments, 5 × 10^5 IKAROS-degron or
   IK6-degron tumor cells were treated with 10 µM asunaprevir or DMSO for
   7 days. After that, asunaprevir-treated cells were washed, and one-half
   were cultured in completed media supplemented with DMSO (ASU/DMSO
   condition), while the other half remained in media supplemented with
   10 µM asunaprevir (ASU/ASU condition). DMSO-treated cells were
   maintained in media supplemented with DMSO (DMSO condition). After 9,
   14, 21, and 28 days, CD19, CD22, and IKAROS levels were measured by
   flow cytometry.

CAR T cell production

   Leukopaks from healthy donors were purchased from StemCell
   Technologies. Primary human T cells were purified by negative selection
   using the RosetteSep Human T cell Enrichment kit (StemCell
   Technologies) and SepMate-50 tubes. T cells were cryopreserved at
   2 × 10^7 cells per ml in CryoStor CS10 cryopreservation medium
   (StemCell Technologies) until use. Isolated T-cells were activated and
   transduced with 19.BBz, 19.28z, 22.BBz, or 19/22.BBz CARs expressing
   retrovirus as described previously^[326]77. Unmanipulated (Mock) cells
   were used as control. CAR T-cell killing in vitro assays were performed
   on day 10 after activation.

IncuCyte tumor killing assays

   IKAROS-degron and IKAROS-degron CD19KO-FL models were treated for 7
   days with 10 µM asunaprevir or DMSO. On Day 0, tumor cells were washed
   and set in co-cultured with mock, blinatumomab-treated (50 ng/ml,
   Invivogen), 19.BBz, 19.28z, 22.BBz, or 19/22.BBz CAR T cells. 5 × 10^4
   GFP-labelled tumor cells was cocultured with 5, 1, or 0.5 × 10^4 CAR T
   cells (E:T 1:1, 1:5, 1:10) in 200 μl RPMI supplemented with 10% FBS,
   10 mM HEPES, 2mM l-glutamine, 100 U ml−1 penicillin and 100 μg ml−1
   streptomycin. Triplicate wells were plated in 96-well flat-bottom
   plates for each condition. Tumor fluorescence was monitored every 2–3 h
   with a ×10 objective using the IncuCyte S3 Live-Cell Analysis System
   (Sartorius), housed in a cell culture incubator at 37 °C and 5% CO2,
   set to take 4 images per well at each timepoint. The total integrated
   GFP intensity was quantified using the IncuCyte basic analyzer software
   feature (IncuCyte S3 v.2019B Rev2; Sartorius). Data were normalized to
   the first timepoint and plotted as the fold change in tumor
   fluorescence over time.

Statistics and reproducibility

   Data were analyzed and visualized using R statistical software
   ([327]http://www.r-project.org) or GraphPad Prism software. The P
   values were calculated with the statistical test described in the
   relevant figure legend. In all cases, a p-value  <  0.05 was considered
   statistically significant. No statistical method was used to
   predetermine the sample size. No data were excluded from the analysis.
   The experiments were not randomized, and the investigators were not
   blinded to allocation during experiments and outcome assessment.

Reporting summary

   Further information on research design is available in the [328]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [329]Supplementary Information^ (9.3MB, pdf)
   [330]41467_2025_58868_MOESM2_ESM.docx^ (13.6KB, docx)

   Description of Additional Supplementary Information
   [331]Supplementary Data 1^ (120.5KB, xlsx)
   [332]Supplementary Data 2^ (41.8KB, xlsx)
   [333]Supplementary Data 3^ (459.6KB, xlsx)
   [334]Supplementary Data 4^ (136KB, xlsx)
   [335]Reporting Summary^ (111.4KB, pdf)
   [336]Transparent Peer Review file^ (16.3KB, pdf)

Source data

   [337]Source Data^ (715KB, xlsx)

Acknowledgements