A Novel Human Fetal Model of CRISPR-Cas9-Induced MLL-AF4 Leukemia Faithfully Recapitulates Infant ALL and Can be Used to Unravel MLL-AF4 Function

Introduction

Infant Acute Lymphoblastic Leukemia (iALL) remains a disease with a dismal prognosis. The most common genetic driver of iALL is a chromosomal translocation that fuses MLL in-frame with AF4, leading to expression of the MLL-AF4 fusion gene and, in ~80% of cases, co-expression of the reciprocal AF4-MLL fusion gene. The translocation invariably arises in utero and appears to be sufficient to drive aggressive iALL without the need for co-operating mutations. Moreover, fetal-specific genes have been detected in iALL blasts and have been implicated in tumor initiation, suggesting developmental stage-associated gene expression programs may contribute to the distinct biology of iALL. Although understanding the mechanisms by which MLL-AF4 initiates and maintains leukemia in a human fetal cell context is crucial for deciphering the biology of iALL, there is currently no human fetal model that faithfully recapitulates MLL-AF4 iALL.

Methods

CRISPR-Cas9 induced double-strand breaks at the MLL and AF4 breakpoints produced a translocation in primary human fetal liver (FL) CD34+ cells. Edited cells and mock-edited controls were (a) cultured on MS5 stromal cells or (b) transplanted into NSG mice that were monitored for leukemia development. Leukemias were characterized by histopathology, immunophenotyping, FISH, IgH rearrangement analysis, RNA-seq and ChIP-seq. The TetR system was used for analysis of MLL-AF4 protein complex interactions.

Results

CRISPR-Cas9-induced MLL-AF4 translocation in FL cells (^CRISPRMLL-AF4+) caused a dramatic increase in B cell proliferation in vitro, with CD19+ cell numbers 900-fold higher in ^CRISPRMLL-AF4+ cultures compared to controls by week 3. qPCR confirmed expression of both MLL-AF4 and AF4-MLL in ^CRISPRMLL-AF4+ cells. Mice transplanted with ^CRISPRMLL-AF4+ cells developed B-ALL (median latency 18 weeks), whereas no control mice developed leukemia. ^CRISPRMLL-AF4+ mice recapitulated key iALL features, including circulating blasts, tissue infiltration and central nervous system (CNS) involvement; and most showed a clonal CD19+CD10- proB immunophenotype. B-ALL was propagated with reduced latency in secondary and tertiary recipients. ^CRISPRMLL-AF4+ ALL closely resembled MLL-AF4 patient samples by RNA-seq and ChIP-seq.

Using publicly available RNA-seq, we derived a human FL (vs adult bone marrow) gene signature and an iALL (vs pediatric ALL) gene signature in order to test whether overlapping onco-fetal gene expression programs were retained in ^CRISPRMLL-AF4+ ALL. At a transcriptomic level, ^CRISPRMLL-AF4+ ALL closely matched poor prognosis HOXA-ve iALL and retained a distinct fetal gene expression signature. Moreover, onco-fetal genes, such as LIN28B, were directly bound by MLL-AF4 in ^CRISPRMLL-AF4+ blasts, suggesting that a human fetal cell context is permissive for MLL-AF4 to bind and retain expression of these genes.

Increased levels of H3K79me2/3 are a hallmark of MLL-AF4 ALL. However, the mechanism by which MLL-AF4 could recruit the H3K79 methyltransferase, DOT1L, is unknown. Analysis of primary human FL cells before and after ^CRISPRMLL-AF4-induced leukemic transformation confirmed that H3K79me2/3 was increased at most MLL-AF4 target genes. However, this chromatin mark alone could not account for the profound changes observed in gene expression. This suggests that other unknown factors may be contributing to both gene regulation and increased H3K79me2/3 in MLL-AF4 ALL. In this study, we show that the protein complex assembled by the AF4 fusion partner is capable of recruiting PAF1 to chromatin, and that anchoring PAF1 to chromatin is sufficient to recruit DOT1L and induce H3K79me2/3, revealing a previously unrecognized role for PAF1 in MLL-AF4 function.

Conclusions

We have developed the first primary human FL derived MLL-AF4+ B-ALL model that expresses both MLL-AF4 and AF4-MLL. ^CRISPRMLL-AF4+ ALL displayed aggressive features characteristic of iALL, including CNS infiltration; and molecular analyses confirmed that it faithfully recapitulates poor prognosis HOXA-ve iALL. Our data suggest that a human fetal cell context plays a key role in the phenotypic and genotypic features of MLL-AF4+ iALL. We also show proof of principle evidence that our fetal ^CRISPRMLL-AF4+ model is a powerful tool for addressing mechanistic questions about MLL-AF4 function and provides an important platform for identifying and testing novel therapies.