AUTHORS
Jannaldo Nieves-Salva, BS, Francisco Gomez-Rivera2, Sarah Khan, M.S.3, Anjali Sengar, PhD4, Karan Bedi, MS5, Jason Xu, M.D./Ph.D.6, Zeel Shukla, B.S.7, Yufei Fu8, Elizabeth Choe9, Petri Polonen, PhD10, Alessandro Baldi, PhD11, Torsten Haferlach, MD, PhD12, David T. Teachey, MD13, Charles G. Mullighan, MBBS (Hons), MSc, MD14, Sethuramasundaram Pitchiaya, Ph.D.15, Mark Chiang, MD, PhD16
AFFILIATIONS
1 University of Michigan, Ann Arbor, Michigan, United States, University of Michigan, Ann Arbor, Michigan, United States
2 University of Michigan, Ann Arbor, Michigan, United States
3 University of Michigan, Ann Arbor, Michigan, United States
4 University of Michigan, Ann Arbor, Michigan, United States
5 University of Michigan, Ann Arbor, Michigan, United States
6 Stanford Health Care, Palo Alto, California, United States
7 University of Michigan, Ann Arbor, Michigan, United States
8 University of Michigan, Ann Arbor, Michigan, United States
9 University of Michigan, Ann Arbor, Michigan, United States
10 St. Jude Children's Research Hospital, Memphis, Tennessee, United States
11 Munich Leukemia Laboratory, Greater Munich Metropolitan Area, Germany
12 Munich Leukemia Laboratory, Greater Munich Metropolitan Area, Germany
13 Childrens Hospital of Philadelphia, Philadelphia, Pennsylvania, United States
14 St. Jude Children's Research Hospital, Memphis, Tennessee, United States
15 University of Michigan, Ann Arbor, Michigan, United States
16 University of Michigan, Ann Arbor, Michigan, United States
ABSTRACT
Recent genomic studies of T-ALL patients from the COG AALL0434 clinical trial identified high-risk subtypes, ETP-like and bone marrow progenitor (BMP)-like, that resemble early T-cell precursors (ETPs) in the thymus. These subtypes are unified by a diverse mutational landscape that converges on stem cell gene expression programs, associating with higher rates of chemoresistance than other forms of T-ALL. Thus, new therapies are needed. To address this, our goal is to dissect transcriptional networks that determine the ETP state. Therapies that shut down the ETP network might be more efficient than individually targeting rare genetic alterations. Towards this goal, we recently reported that the PIAS-like coactivator ZMIZ1 induces an essential native transcriptional network of coherent feedforward circuits of ETP enhancers and oncogenes (PMID 39969525). This network helps define the ETP state and contributes to cardinal features. A separate group recently reported in AML cells that ZMIZ1 forms enhancer-associated foci with biophysical and biochemical properties of liquid-like droplets, which we confirmed in ETPs. These data raise the intriguing possibility that ZMIZ1 organizes enhancer condensates.
In order to learn how to inhibit ZMIZ1, we examined leukemia genomes for potential regulatory sequences undergoing selection. We identified somatic ZMIZ1 mutations enriched in AALL0434 ETP-like cases. While these mutations are rare in children/AYA (~2%), the MLL dataset shows that mutation frequency rises to ~18% in adult ETP-like. ASHOP analysis showed that ZMIZ1 expression is 4-fold lower in adult vs pediatric ETPs, suggesting selection for mutation when ZMIZ1 dosage is limiting. The mutations cluster in the alanine-rich domain (ARD, AA280-305), which has no known function. In a neurodevelopmental (neuro) disorder, germline ARD mutations are loss of function. However, ZMIZ1-mutant AALL0434 cases expressed ZMIZ1 target genes ~30% more highly than ZMIZ1-wildtype cases, suggesting that T-ALL mutations are gain-of-function. To test this, we generated a Cre-activated mouse model of the Δ293-296 mutation. Mutant mice showed ~2-fold increased ETP cell number and impaired differentiation compared to controls. Next, we tested the effect of mutations on condensate formation in U2OS cells at ectopic, comparable, ETP-equivalent expression levels. Immunofluorescence (IF) microscopy showed that mutants increased number of condensates 30-32% and partition percentage 23-25% compared to wildtype. In addition, mutant condensates co-enriched for enhancer biomolecules to 1.5-2.2-fold higher than wildtype. Next, we transduced mutants into a T-ALL cell line. Mutant ZMIZ1 protein expression was 1.5-2.7-fold higher than wildtype by IF, flow cytometry, and Western blot. mRNA was unchanged. Mutants showed 3.6-4.5-fold increased Myc, a ZMIZ1 target gene, and ~1.3-fold increased cell growth compared to wildtype. In contrast, the neuro mutant T300M showed reduced protein expression, consistent with prior reports. Thus, T-ALL and neuro mutants confer opposite functional properties.
Next, we wondered whether the ARD has roles in protein-protein interactions and biomolecular condensation since it is intrinsically disordered. To test this, we deleted the ARD in T-ALL cells. ZMIZ1 protein expression increased ~2-fold, suggesting the ARD is a degron. Consistently, cycloheximide chase assay showed that ΔARD and T300M increased and decreased protein stability, respectively. When fused to GFP ("GFP-ARD"), the ARD reduced protein stability 100-200-fold, which was partially reversed by proteasome inhibition or T-ALL mutation. Mutation increased condensate formation compared to wildtype. Next, we performed mass spectrometry of GFP-ARD chimera pulldowns. Wildtype GFP-ARD showed strong differential binding to proteasome and chromatin proteins relative to mutant or GFP alone.
Taken together, these data suggest that ZMIZ1 molecules exist in dynamic equilibrium between a transcriptionally active, condensate-associated state and an unstable, proteasome-associated state. Wildtype and neuro mutant ARDs shift this equilibrium to the unstable state, while opposing T-ALL mutants push it to the active state. Thus, the ARD might be a peptide signal that tunes enhancer activity bidirectionally through coordinate regulation of compartment partitioning and protein concentration. In theory, our data suggests that ETP leukemias can be targeted by driving ZMIZ1 to the unstable state.
DISCLOSURES
Jannaldo Nieves-Salva, BS: No relevant conflicts of interest to declare.