IKAROS levels are associated with antigen escape in CD19- and CD22-targeted therapies for B-cell malignancies

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.

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

A CART19 cohort of PDX samples analyzed by CyTOF. Created in BioRender. 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).

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

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. 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^, (n = 11), adult B-ALL patients treated with blinatumomab (n = 2), and adult LBCL patients treated with 19.28z CAR T cells^, (n = 6) (H). I GSEA results for Zheng Cord Blood C6 HSC/MPP gene signature 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. 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.

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

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).

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

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).

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

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.