Genomic determinants of response and resistance to inotuzumab ozogamicin in B-cell ALL

Abstract

Inotuzumab ozogamicin (InO) is an antibody-drug conjugate that delivers calicheamicin to CD22-expressing cells. In a retrospective cohort of InO-treated patients with B-cell acute lymphoblastic leukemia, we sought to understand the genomic determinants of the response and resistance to InO. Pre- and post-InO-treated patient samples were analyzed by whole genome, exome, and/or transcriptome sequencing. Acquired CD22 mutations were observed in 11% (3/27) of post-InO-relapsed tumor samples, but not in refractory samples (0/16). There were multiple CD22 mutations per sample and the mechanisms of CD22 escape included epitope loss (protein truncation and destabilization) and epitope alteration. Two CD22 mutant cases were post-InO hyper-mutators resulting from error-prone DNA damage repair (nonhomologous/alternative end-joining repair, or mismatch repair deficiency), suggesting that hypermutation drove escape from CD22-directed therapy. CD22-mutant relapses occurred after InO and subsequent hematopoietic stem cell transplantation (HSCT), suggesting that InO eliminated the predominant clones, leaving subclones with acquired CD22 mutations that conferred resistance to InO and subsequently expanded. Acquired loss-of-function mutations in TP53, ATM, and CDKN2A were observed, consistent with a compromise of the G1/S DNA damage checkpoint as a mechanism for evading InO-induced apoptosis. Genome-wide CRISPR/Cas9 screening of cell lines identified DNTT (terminal deoxynucleotidyl transferase) loss as a marker of InO resistance. In conclusion, genetic alterations modulating CD22 expression and DNA damage response influence InO efficacy. Our findings highlight the importance of defining the basis of CD22 escape and eradication of residual disease before HSCT. The identified mechanisms of escape from CD22-targeted therapy extend beyond antigen loss and provide opportunities to improve therapeutic approaches and overcome resistance. These trials were registered at www.ClinicalTrials.gov as NCT01134575, NCT01371630, and NCT03441061.

Graphical abstract

Figure 1.

Response to inotuzumab. (A) Swimmer plot for InO responders (n = 51). Each bar represents the start of the InO therapy to the last follow-up. Once attaining CR/CRi, patients were off InO therapy, and 31 responders subsequently received HSCT. (B) Flow chart summarizing the types of samples and the outcome of the patients. CR, complete remission; CRi, CR with incomplete hematologic recovery; HSCT, hematopoietic stem cell transplantation.

Figure 2.

Multiple mechanisms of CD22 antigen escape. (A) Protein domain plot of all the post-InO–acquired CD22 mutations (n = 10) identified in this cohort. (B) WT and predicted mutant CD22 protein structures. Group by mutation type. (C) Clinical flow cytometry plots of paired pre- and post-InO patient samples. CD22 status (WT or mutations) was labeled below each flow plot. The relative abundance of alternative splicing of CD22 e1 was depicted as stacking plots at the bottom. e1_e2 was the canonical splicing junction, and e1_e3, e1_e3alt, e1_e4, e1_e5, e1_e6, and e1_e7 were alternative splicing junctions. (D) Antibodies directed to different CD22 epitopes. InO and SHCL1 that bind the first Ig domain of CD22, and RFB4 that bind the third. (E-F) Functional characterization of CD22 p.R75T and p.C529Y. CD22 protein expression (by western blot) and binding (by flow cytometry) were performed using antibodies directed to different CD22 epitopes. All controls and CD22 mutants were run in the same experiment of western blot or flow cytometry. The results for CD22 p.R75T and CD22 p.C529Y were presented in seperate panels with the same controls (EV, CD22 WT). (G) Stack plots depicting the relative abundance of alternative splicing of CD22 e1 in pre-InO baseline samples, comparing responders (EFS >12 m and EFS <12 m) and nonresponders. (H) Stack plots depicting the relative abundance of alternative splicing of CD22 e1, comparing pre- and post-InO samples. EV, empty vector; WT, wild-type; m, month.

Figure 3.

TP53 mutations in primary and acquired resistance to InO. (A) Baseline genomic alterations in major leukemia genes, comparing responders and nonresponders. (B) EFS and overall survival in TP53WT vs TP53mut patients. The elapsed observation time is plotted on the curve as a circle. Symbols for both events and censored observations were plotted so that each subject was shown. The censored observations can be clearly seen as circles along the horizontal portion of the curve. The medians were estimated using the Kaplan-Meier method. P values were determined using the log-rank test. (C) Dynamic changes in TP53 VAF in 5 cases with paired pre- and post-InO samples. The VAF for each mutation is shown along with the sample blast count at the corresponding time point. The time elapsed from the start of InO to treatment failure (R/R) was is in days. CI, confidence interval; HR, hazard ratio; mut, mutation; NGS; next-generation sequencing; WES, whole-exome sequencing; WGS, whole-genome sequencing; WT, wild-type.

Figure 4.

Post-InO–acquired determinants of resistance. (A) Number (top plot) and OncoPrint (bottom plot) of post-InO–acquired mutations. For each patient, acquired mutations were mutations present in post-InO but not in pre-InO. The dotted lines (at 300 and 900, number of post-InO–acquired mutations) were used as cutoff to divide patients into post-InO low-mutators, moderate-mutators, and hyper-mutators. (B) Fish plots showing acquisition of mutations in CD22 and KMT2D. (C) Mutational signatures of post-InO–acquired mutations in the hyper-mutators SJALL058834 (>10 bp insertions) and SJALL074541 (SBS6 and MMR-deficient). (D) A simplified schematic of the compromised G1/S DNA damage checkpoint by acquired mutations in InO-treated patients. SBS, single base substitution; DBS, doublet base substitution; HSCT, hematopoietic stem cell transplantation; NR, nonresponder. Figure created with BioRender.com.

Figure 5.

Genome-wide CRISPR/Cas9 screening with InO in the NALM-6 cell line (passage 10). (A) Volcano plots showing InO sensitization (negatively selected, in blue) and resistance (positively selected, in red) genes by CRISPR/Cas9 screening. (B) GSEA of negatively enriched pathways. Magma colors and size of the dots represent the normalized enrichment score (NES) and −log₁₀ false discovery rate (FDR) q-values of the top 20 negatively enriched pathways, respectively. The respective gene set members of each pathway are denoted by a bubble plot. Viridis colors and size of the dots represent running enrichment score and −log₁₀ of CRISPR/Cas9 P value respectively. (C) Log₂ FC values of sgRNAs targeting the top 10 InO resistance genes. Each gray line represents an individual sgRNA and the central dotted line represents the median log₂ FC of the sgRNAs (∼0). Each red line represents the sgRNA of the target gene. Multiple sgRNAs targeting each of the top 10 genes (including CD22) were significantly enriched by inotuzumab, showing an on-target gene knockout by CRISPR/Cas9. (D) DNTT RNA expression in paired patient samples compared with pre- and post-InO. The plot represents the mean ± standard error of the mean for each group. P = .03 by two-tailed paired t test. Samples belonging to the same patient were connected using lines. CPM, counts per million FC, fold change; GSEA, gene set enrichment analysis; sgRNAs, single-guide RNAs.