Paper Details

ABSTRACT

Background:

Blinatumomab is transforming the treatment of childhood B-cell acute lymphoblastic leukaemia (B-ALL), improving efficacy while reducing exposure to toxic chemotherapy. The biological basis of disease that is resistant to blinatumomab is unclear. We aimed to define lymphoblast and immune-cell features associated with non-response and determine whether they changed with the emergence of resistant disease.

Methods:

We contextualised children treated with blinatumomab within a single-cell compendium spanning childhood leukaemia subtypes. The cohort included children treated for persistent B-ALL with detectable lymphoblasts before treatment. Longitudinal bone-marrow samples underwent single-cell mRNA sequencing, supplemented by immune-receptor and whole-genome sequencing; available cDNA libraries underwent long-read sequencing. Response required MRD clearance and remission for at least 2 years; non-response comprised refractory disease or relapse after molecular remission. Blast transcriptional states were compared cross-sectionally and in paired blinatumomab-naive and resistant samples. We also analysed non-malignant immune cells, routine clinical flow cytometry from an expanded cohort and PBMC–B-ALL co-cultures with blinatumomab and JAK1/2 inhibition.

Results:

The childhood leukaemia compendium comprised 2,476,676 cells from 161 individuals and 311 samples. We focused on 102 longitudinal samples from 37 children treated with blinatumomab with persistent disease prior to therapy: 18 achieved sustained remission and 19 were non-responsive. Non-response occurred across 9 genetic subtypes. No clinical, laboratory or recurrent genetic features readily distinguished responders from non-responders.

Instead, complementary transcriptional analyses converged on a lymphoblast phenotype of diminished immune receptivity. Principal-component analysis separated responsive and non-responsive samples along an axis characterised by reduced B-cell precursor and HLA class II gene expression in non-responders. Non-negative matrix factorisation identified significant depletion of B-cell identity and antigen-presentation gene co-clusters in non-responsive lymphoblasts; inferred transcription-factor activity similarly showed reduced activity of interferon-regulatory factors. Gene-set scoring showed that this state was already detectable before treatment: blinatumomab-naive non-responders had lower MHC class I and II (both q=0.007) and IFN-α and IFN-γ (both q=0.04) module scores than responders. These programmes remained low in resistant disease; in 9 of 10 paired cases, lymphoblasts shifted further along the same transcriptional axis after exposure (P=0.037), demonstrating further accentuation at resistance.

Targeted flow cytometry of primary patient samples confirmed lower surface HLA class II expression. CD19 transcript and surface expression were not significantly reduced, and CD19 loss was not observed at failure. Long-read sequencing (n=31 samples) detected abnormal CD19 isoforms, including exon 2 skipping and intron 2 and/or 6 retention, at similar frequencies in responders and non-responders.

We examined 305,130 non-malignant immune cells across 82 longitudinal B-ALL samples. Immune-cell composition, T-cell clonal expansion and T-cell transcriptional states during blinatumomab were similar between response groups. In an expanded clinical flow-cytometry cohort of 89 children, persistence of normal CD19-positive B cells during treatment was enriched among non-responders (P=0.0053) and remained associated after adjustment for detectable disease.

In vitro studies across 2 B-ALL cell lines showed that blinatumomab induced dynamic HLA class I and II upregulation alongside T-cell activation and lymphoblast killing; JAK1/2 inhibition diminished each response, demonstrating that JAK1/2-dependent signalling contributes to HLA induction and effective blast–T-cell interactions.

Conclusions:

Non-responsive lymphoblasts converge on reduced HLA expression and attenuated interferon-associated signalling, frequently detectable before blinatumomab and accentuated in resistant disease. These findings identify diminished lymphoblast immune receptivity as a cellular basis of non-response, nominate blast HLA expression and on-treatment normal B-cell depletion as measurable correlates, and provide a framework for predicting and overcoming resistance as blinatumomab enters frontline childhood B-ALL therapy.