Oral Presentation and Poster Abstracts

Presenter: Nicole Michmerhuizen, PhD

Mentor: Charles Mullighan, MBBS, MD

Co-Authors: Emily B. Heikamp, Ilaria Iacobucci, Masayuki Umeda, Bright Arthur, Danika Di Giacomo, Ryan Hiltenbrand, Qingsong Gao, Pablo Portola, Aurelie Claquin, Bappaditya Chandra, David W. Baggett, Ali Khalighifar, Hongling Huang, Peipei Zhou, Lingyun Long, Hao Shi, Yu Sun, Evangelia K. Papachristou, Chandra Sekhar Reddy Chilamakuri, Francisca N. de L. Vitorino, Joanna M. Gongora, Laura Janke, Sandi Radko- Juettner, Josi Lott, Lei Yang, Xiang Fu, Mary Ashley Rimmer, Kimberly S. Mercer, Alex Kentsis, Clive S. D’Santos, Benjamin A. Garcia, Richard W. Kriwacki, Hongbo Chi, Jeffery M. Klco, Scott A. Armstrong, Charles G. Mullighan

The Histone Acetyltransferase MOZ is a Molecular Dependency and Therapeutic Target in NUP98-Rearranged Acute Myeloid Leukemia

Abstract
NUP98 fusion oncoproteins (FOs) are a hallmark of childhood acute myeloid leukemia (AML), and drive leukemogenesis through liquid-liquid phase separation-mediated nuclear condensate formation. However, the composition and consequences of NUP98 FO-associated condensates are incompletely understood. To identify FO interacting proteins, we performed rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME) for wildtype NUP98 and eight NUP98 FOs (NUP98::HOXA9, PMX1, KDM5A, NSD1, JADE2, LNP1, RAP1GDS1, SETBP1) in transfected HEK293T cells. In addition to known NUP98 FO interactors (e.g. XPO1, RAE1), we identified novel interactors including MOZ and other histone acetyltransferase (HAT) complex members. Further, in vivo CRISPR/Cas9 screening with a library of epigenetic gRNAs suggested that BRPF1, an epigenetic writer that associates with MOZ and other HAT complexes, is a molecular dependency in NUP98::KDM5A AML. Competitive co-culture of Nup98::Kdm5a;Vav-Cre cells showed that inactivation of Brpf1 as well as Moz, Hbo1, Brd1 or Meaf6 decreased NUP98-r cell fitness. MOZ inhibition decreased global H3K23ac levels, displaced FO from chromatin at the Meis1 locus, and led to myeloid cell differentiation in cells from human and mouse models of NUP98 rearrangement. MOZ inhibition also decreased leukemic burden in multiple NUP98-rearranged leukemia xenograft models. Finally, we identified that treatment with MOZ inhibitor synergized with Menin inhibitor treatment in vitro and in vivo. Combination treatment downregulated NUP98 FO target genes, including many in the Hoxb cluster, and heightened myeloid differentiation beyond what was observed with single agent therapy. Importantly, combination therapy extended survival in a highly refractory and Menin inhibitor-resistant NUP98::NSD1 patient-derived xenograft. In summary, we show that MOZ is a potentially targetable dependency in NUP98-rearranged AMLs.

Presenter: Keren Zhou, PhD

Mentor: Ruth Tatevossian, MD, PhD

Co-Authors: Sujuan Jia, Wenchao Zhang, Karol Szlachta, Yiping Fan, Ti-Cheng Chang, Gang Wu, Ruth G. Tatevossian, Charles G. Mullighan

Investigating the potential of cell free RNA as a biomarker in fusion positive childhood B-cell acute lymphoblastic leukemia

Abstract
Introduction: B-cell acute lymphoblastic leukemia (B-ALL) is the most prevalent subtype of childhood leukemia. Aneuploidy or oncogenic translocations define leukemia subtypes of prognostic significance to initiate leukemogenesis. Plasma cell-free RNA (cfRNA), originating from circulating cells, represents a wide range of RNA species including messenger RNAs (mRNAs), which retain the molecular characteristics of liquid tumors. However, cfRNA is typically highly fragmented and present in low concentrations, making detection and characterization challenging. We sought to investigate the potential utility of cfRNA as a biomarker in fusion positive childhood B-ALL.

Methods: We extracted cfRNA from plasma samples collected at diagnosis from children with B-ALL. Oncogenic fusions were confirmed in all cases. Stranded libraries were prepared using an exome enrichment kit and submitted for high- throughput sequencing via Illumina and Nanopore platforms. We analyzed gene expression profiles, chimeric fusion transcripts, chromosomal level copy number changes, and other key molecular characteristics, such as single nucleotide variants (SNVs) and insertions/deletions (indels), from the resulting sequencing data.

Results: Leveraging the power of long-read sequencing, we revealed that there was a wide distribution of cfRNA fragment sizes from approximately 30 bp to 5 kb. We observed strong correlations between gene expression profiles from cfRNA and corresponding primary tumor cell RNA. Using a machine learning classifier well-trained on the bulk RNA- seq from primary tumor cells from children B-ALL cohort, we achieved high accuracy in B-ALL subtyping by utilizing the gene expression profiles derived from cfRNA. Chromosome level copy number changes identified from cfRNA resemble the corresponding primary tumor bulk RNA-seq. Specific chimeric fusion transcripts including TCF3::PBX1, KMT2Ar, ETV6::RUNX1, BCR::ABL1 and IGH::DUX4 were also successfully detected from cfRNA. Furthermore, we identified key SNV and indels in cfRNA, which were well characterized in corresponding WGS data.

Conclusions: cfRNA faithfully preserves critical information from RNA molecules in tumor cells of pediatric B-ALL patients and shows significant potential as sensitive biomarkers for the diagnosis and monitoring of pediatric B-ALL.

Presenter: Paula Arnold, PhD

Co-Authors: Subodh Selukar, Guangjin Luo, Ying Li, Brandon Triplett, and Swati Naik

Relapse of pediatric acute leukemias with genomic HLA loss following haploidentical hematopoietic cell transplant

Abstract
Immune cells in the hematopoietic cell transplant (HCT) graft provide graft-versus-leukemia effects but immune pressure may lead to tumor escape and relapse. Loss of recipient-specific human-leukocyte antigen (HLA) genes occurs in 30-50% of relapses following haploidentical (haplo)HCT in adults but studies in pediatric patients are limited. Testing for HLA loss after HCT relapses may inform clinical management including benefit of donor lymphocyte infusion (DLI) or alternative donor selection for rescue HCT.

We performed retrospective analysis of 53 pediatric patients with acute leukemias who underwent 60 haploHCT at our institution and had post-HCT relapse samples tested for genomic HLA loss. We summarized baseline and post-HCT characteristics descriptively by relapse classification, either as ‘classical’ or ‘HLA loss’ relapse. We estimated the overall survival (OS) probability from time of HLA loss relapse until all-cause mortality or last contact date with the Kaplan-Meier method. We used logistic regression to evaluate risk factors for relapses that accompanied HLA loss.

Recipient-specific HLA loss was noted in 25% of post-haploHCT relapses. The odds of relapsing with HLA loss were higher in patients who entered transplant with active disease (>5% blasts) or who were in complete remission (CR)3/4, relative to patients transplanted in CR1/2 (p= 0.035). There was a trend for higher odds of HLA loss relapse in patients who developed acute graft versus host disease (aGVHD, p=0.077). There was no significant difference in the frequency of HLA-loss relapses compared to classical relapses based on leukemia type, number of HLA mismatches in the GVH direction, T-replete or T-deplete grafts, number of T cells in the graft, time from HCT to relapse, or number or dose of prophylactic DLI. Overall survival after HLA loss relapse at 1-year and 3-year was 30.0% and 22.5%, respectively.

The rate of HLA loss in pediatric haploHCT recipients was comparable to reports in adult patients and was associated with poor survival. Factors associated with susceptibility to immune pressure, including disease status at HCT and occurrence of aGVHD were associated with incidence of HLA-loss relapses. Given frequency and clinical implications, we suggest routine testing for HLA-loss for relapses post-haploHCT in pediatric patients.

Presenter: Selene Koo, MD, PhD

Mentor: Brent Orr, MD, PhD

Co-Authors: Agda Laakso, Quynh Tran

MGMT promoter methylation and MGMT staining status in pediatric solid tumors

Abstract
O6-Methylguanine-DNA Methyltransferase (MGMT) encodes a key DNA repair enzyme. In glioblastoma, MGMT promoter methylation (MGMT-PM) predicts poor progression-free survival and increased sensitivity to temozolomide, an alkylating chemotherapeutic agent. However, its role in non-brain tumors is not well studied. We studied MGMT-PM status in a large cohort of pediatric solid tumors.

MGMT-PM status was determined using Illumina Infinium methylation array data from 4695 patient tumors and 265 tumor cell lines. MGMT immunohistochemical stain (IHC) was performed on matched formalin-fixed, paraffin-embedded tissue sections. Stain was considered retained if >20% of tumor cell nuclei were positive and lost if <20%. Scoring was performed blinded to MGMT-PM.

Tumor types with the highest frequency of MGMT-PM were pediatric adrenocortical tumor (5.6%), osteosarcoma (OS; 4.5%), and rhabdomyosarcoma (RMS, 2.9%), including alveolar RMS (ARMS; 4.9%), embryonal RMS (ERMS; 2.0%), and spindle cell/sclerosing RMS (4.5%). Among cell lines, the highest MGMT-PM frequency was found in RMS (54.5%; 90% in ARMS and 25% in ERMS), OS (50%), and neuroblastoma (10.2%). In 39 RMS patients with multiple tumor samples, MGMT-PM status was consistent across all samples.

MGMT IHC was performed on a subset of RMS cases (n=38; 7 MGMT methylated and 31 MGMT unmethylated). Five MGMT-methylated cases had concordant MGMT IHC loss, 2 MGMT-methylated cases had discordant retention of MGMT IHC, and all 31 MGMT- unmethylated cases had retention of MGMT IHC, for sensitivity of 0.71 and specificity of 1.

MGMT-PM is rare but more common in certain pediatric solid tumors, such as OS and ARMS. This suggests potential therapeutic options, such as alkylating agents, in treatment- refractory tumors. Preliminary data suggests that in pediatric RMS, MGMT IHC may be a useful screening tool for assessing MGMT-PM status.

Presenter: Juan M. Barajas, PhD

Mentor: Jeffrey Klco, MD, PhD

Co-Authors: Aaron Phillips, Jina Wang, Melvin Thomas III, Lisett Contreras, William T. Freyaldenhoven, JingSong Fang, Kelly Churion, Ryan Hiltenbrand, Elizabeth Caldwell, Katelyn Jackson, Masayuki Umeda, Tamara Westover, Emily Xiong, Josi Lott, Sandi Radko-Juettner, Wenjie Qu, Beisi Xu, Jing Ma, Laura Janke, Richard Kriwacki, Jeffery M. Klco

Nuclear export signals in UBTF tandem duplications are necessary for leukemic transformation

Abstract
Recurrent in-frame tandem duplications in UBTF are a defining genomic alteration in myeloid neoplasms. First described in pediatric acute myeloid leukemias (AML), UBTF-TD alterations have now been reported in ~3% of adults with AML aged 18-60 years and cases of myelodysplastic syndrome (MDS). These findings underscore the need for a deeper understanding of UBTF-TD-associated myeloid neoplasms. The UBTF gene encodes a protein crucial for ribosomal RNA expression and nucleoli formation. However, the molecular function of UBTF- TD in leukemic transformation remains poorly understood. 

Recently, we identified XPO1 as a novel interacting partner unique to UBTF-TD and not UBTF-WT using quantitative rapid iummonuprecipitation of endogenous proteins (qPLEX-RIME) which assess on-chromatin protein interactions. We therefore analyzed the amino acid sequences of 97 available cases of UBTF-TD myeloid malignancies to look for the presence of nuclear export signal (NES) motifs. We found that, in 95 out of 97 cases (97.9%), the tandem duplications in UBTF produce a hydrophobic sequence resembling a NES recognized by the nuclear export protein XPO1. Using a cord-blood CD34+ model of UBTF-TD, we find that the introduction of alanine mutations that disrupt the NES sequence led to loss of cellular transformation and a loss of chromatin interaction at functionally relevant regions like HOXA/HOXB gene clusters. We also used qPlex-RIME and GST pulldowns to assess protein interactions of UBTF-TD and UBTF-TD alanine mutants. We show that alanine mutations of the NES sequence decreases the interaction of UBTF-TD with XPO1. These data suggest that UBTF-TD is dependent on the NES sequence for interaction with XPO1 and leukemic transformation. Furthermore, we find that treatment of a UBTF-TD patient derived xenograft (PDX) model with the XPO1 inhibitor Eltanexor results in eviction of UBTF-TD from chromatin and decreased tumor burden. 

In summary, our study sheds light on a potential mechanism whereby UBTF-TD's on-chromatin interactions with XPO1 through a de novo nuclear export signal are essential for leukemic transformation and provide evidence for the use of XPO1 inhibitors in patients with UBTF-TD alterations.

Presenter: Alex Breuer, PhD

Mentor: Lu Wang, MD, PhD

Co-Authors: Larissa Furtado, Brent Orr, Lu Wang

Neural network model builds may not be reproducible: implications for AI in Pathology

Abstract
AI tools have demonstrated transformational potential for pathology. As AI tools are incorporated into clinical pathology workflows, it is essential to discover and establish the limits of their applicability and utility. We concentrate on repeatability—particularly, the capability to reproduce an AI model from train- ing data. This kind of repeatability has significant clinical implications: a classifier that is used clinically may be required to be updated due to software upgrades, and a model that lacks repeatability would present a particular challenge to such upgrades.

We provide a concrete example from methylation classification that shows even simple neural network models may not be reproducible on different hardware even when all training data, model parameters, and software versions are held fixed. We use the familiar central nervous system tumor set from Capper et. al. to train a series of multilayer perceptrons to classify members of the glioblastoma IDH-wild type methylation family. These perceptrons all share the same architecture: 2 hidden layers, implemented in PyTorch, and are intended to test the limits of repeatability. We train 2 replicates of the model on each of 3 different hardware platforms; two GPU platforms and one CPU platform. We use the same ansatz for all model parameters, use the maximum numerical precision possible, and keep all software library versions identical across the compute machinery we use. While the two replicates from each hardware platform are identical, each hardware platform produces its own unique neural network, even with the ansatz and all software and library versions fixed. Moreover, these three neural networks produce mean- ingfully different results for some cases—some neural nets produce clinically-usable calls while others do not for the same case. Depending on the clinical threshold used, it is even possible to encounter cases which produce clinically reportable results that differ between the neural networks. We contrast this ex- perience with two replicates of logistic-calibrated kernel ridge regressors trained on the same hardware and software as the neural nets; these calibrated kernel ridge regressors produce identical results over all hardware platforms.

It is already well-known that neural networks do not have convex loss functions and that round-off errors can and do differ between different hardware.   We  show that the failure of neural net reproducibility    can be attributed to the combination of round-off error and the non-convexity of the neural network     loss function through an investigation of the curvature of this network’s loss function over iterates of    the fitting algorithm. Conversely, we observe the also well-known fact that kernel ridge regressors calibrated with logistic regression are reproducible up to machine tolerances even where the neural network models diverge. These results can inform choice of model for clinical deployment and software lifecycle considerations.

Presenter: Sarada Achyutuni, PhD

Mentor: Jian Xu, PhD

Co-Authors: Xiaofei Gao, Hieu S Vu, Yuannyu Zhang, Hui Cao, Aditya Sheth, Ilaria Iacobucci, Jeffery Klco, Charles G. Mullighan, Stephen S. Chung, Min Ni, Jian Xu

Identifying and characterizing the metabolic liabilities of leukemia-initiating cells

Abstract
Leukemia-initiating cells (LICs) play a crucial role in leukemia initiation, progression, and relapse. These cells undergo metabolic reprogramming to adapt to the tumor microenvironment, primarily influenced by their oncogenic driver and/or cell-of-origin and exhibit unique metabolic vulnerabilities. Altered glucose, lipid, and amino acid metabolism, along with increased oxidative stress are key metabolic features of LICs. Targeting these LIC-selective metabolic dependencies offers a promising therapeutic strategy. 

To identify LIC-selective metabolic vulnerabilities, we conducted metabolomic and transcriptomic profiling of over 1,000 murine and human leukemia samples, including commonly used cell lines, patient-derived xenografts, and primary leukemia samples. Our analysis identified distinct; leukemia subtype-specific metabolic signatures linked to the underlying genetic drivers. For example, BCR-ABL-positive chronic myelogenous leukemia (CML) relies on oxidative phosphorylation, while MLL-rearranged acute leukemia exhibits altered lipid metabolism. We identified a critical metabolic pathway in RUNX1-RUNX1T1 (AE9a)-driven myeloid leukemia. AE9a-positive LICs showed a strong dependence on glutaminase (GLS), an enzyme crucial for glutamine metabolism. GLS inhibition using CB839 led to cell cycle arrest, elevated reactive oxygen species (ROS), lipid peroxidation, and ferroptosis. Furthermore, GLS inhibition impaired the synthesis of glutathione in AE9a-driven LICs, a major cellular reducing agent, whereas AE9a enhances the expression of GLS and ferroptosis-associated genes, including GPX4, a key regulator of ferroptosis. 

These findings suggest that AE9a-driven LICs rewire intracellular metabolism to promote glutathione biosynthesis and protect cells against lipid peroxidation, whereas targeting GLS and GPX4 could enhance ferroptosis, selectively impairing AE9a-positive LICs. Together, our study highlights the potential of targeting LIC-selective metabolic dependencies across leukemia subtypes as single or combinatorial regimens for improved leukemia therapy.

Presenter: Victor Pastor, PhD

Mentor: Lu Wang, MD, PhD

Co-Authors: Victor B. Pastor, Jamie L. Maciaszek, Enrico Attardi, Sara Lewis, Taylor M. Cain, Passant Shaker, Nathan Gray, Michelle Boals, Jennifer Neary, Michael Rusch, Mark R. Wilkinson, Lu Wang, Jeffery M. Klco, Marcin W. Wlodarski

Integration of a Bone Marrow Failure Germline Panel into the Comprehensive Clinical Genomics Pipeline at St. Jude Children’s Research Hospital: Insights from 18 Months’ Experience

Abstract
Introduction: The diagnosis of bone marrow failure (BMF) syndromes is challenging due to overlapping clinical features with other conditions and time-consuming tests. Advances in genomic testing through comprehensive gene panels now enable rapid and precise diagnosis, improving clinical management.

Methods: Germline BMF testing using whole-genome and exome sequencing targeting 155 BMF- associated genes was implemented at our institution in April 2023. Patients identified by their clinicians as having clinical features of bone marrow failure or related hematologic disorders were selected for genetic testing. Genetic counselors provided pre-test counseling, obtained informed consent, and facilitated interpretation and post-test counseling upon return of germline results. To date, 56 patients have undergone testing, with samples obtained from direct skin biopsies (71%) or blood draws (29%). Additionally, 17 patients with a confirmed or suspected cancer diagnosis underwent complementary testing using a 115- gene cancer predisposition panel.

Results: Initial presentations included 10 BMF syndromes (Diamond-Blackfan anemia, GATA2 deficiency, SAMD9/9L syndromes), 28 aplastic anemia/cytopenia, 9 acute myeloid leukemia (AML), 1 B-acute lymphoblastic leukemia (BALL), 2 myelodysplastic syndromes and 6 treatment related hematologic disorders after solid or brain tumors. Germline testing identified 17 pathogenic variants in 15 patients, including exon-level deletions in FANCA and a whole-gene deletion in RPL15, underscoring the value of comprehensive genomic analysis.

Molecular diagnostics provided confirmatory diagnoses and revealed incidental findings in 9 patients. Three patients, including 2 initially diagnosed with AML and 1 with pancytopenia, were reclassified as Fanconi anemia following the identification of pathogenic FANCA/G variants. One patient, initially diagnosed with pancytopenia, received a confirmatory diagnosis of dyskeratosis congenita following the identification of a novel homozygous variant in the CTC1 gene. Similarly, another patient presenting with cytopenia was confirmed to have MECOM-associated syndrome. Further, a patient with retinoblastoma and delayed count recovery was diagnosed with Bloom syndrome due to a homozygous BLM variant. Incidental findings in 3 patients included variants associated with cancer predisposition in BRCA2, POT1 and CHEK2.

Conclusions: Germline testing refined diagnoses in 11% of patients (6/56), confirming conditions such as Fanconi anemia, dyskeratosis congenita, and Bloom syndrome. Additionally, incidental findings in genes predisposing to cancer were identified in 5% of patients (3/56). Overall, pathogenic variants were detected in 27% of cases (15/56), highlighting the critical role of genomic testing in improving diagnostic precision and guiding personalized care, particularly for patients with complex hematologic and oncologic presentations.

Presenter: Clair Kelley, PhD

Mentor: Priya Kumar, MD, PhD

Co-Authors: Clair Kelley, Zhaohong Yin, Paul Mead and Priya Kumar

Evaluation of Fixatives for Preservation and Immunophenotyping of Pediatric Leukemia Samples for Transport in Limited Resource Settings (LMIC)

Abstract

Introduction: Flow Cytometry (FC) is a powerful tool for rapid and accurate diagnosis of leukemia and MRD evaluation. Due to expense and technical expertise required, patient samples are often sent to a centralized site. Blood samples in ETDA tubes are stable 24-48 hours, but in LMICs, transportation can take days. This can lead to loss of sample integrity required for accurate diagnosis. Fixatives have been developed to preserve cells for FC, but efficacy for analysis of pediatric leukemia samples or temperatures typical in LMICs are unknown. We aim to identify a fixative that will maintain samples for accurate diagnosis by FC.

Methods: We have tested 5 fixatives on samples stored at 24oC for 5, 7 and 10 days. FC analysis was run with an 8 color immunophenotyping (IP) panel containing antibodies for 5 surface and 3 cytoplasmic markers. Leftover pediatric AML or B-ALL samples were evaluated for viability and FC providing a day 0 baseline prior to adding fixatives. At 5, 7 and 10 days, FC was run on fixed and unfixed (UF) samples. Data was evaluated for population separation by CD45/SSC, identification of normal B and T cells and granulocytes and the identification and IP of leukemic blasts. The functionality of each antibody was compared for each fixative and UF at each timepoint.

Results: TFX maintained the best CD45/SSC profile at all time points. Separation of CD45/SSC populations in samples fixed in NAB or SCP were poor and samples fixed with GT or CHX were comparable to UF. In all conditions cyCD3/sCD3 worked well in T cell identification. Of the B cell markers, CD19 and CD79a worked better in all fixatives than UF, except that CD79a was lost in TFX. Blasts could be identified in all leukemia samples through day 10, but % identified varied by immunophenotype and fixative. Across all conditions, CD10 on blasts was maintained well and identification of blasts based on CD10 positivity was close to that of UF D0. CD34 and MPO also worked well in blast identification, with somewhat lower performance of CD34 in NAB and of MPO in TFX. Given that TFX worked best except for cD79a, we will try other B markers, such as CD22.

This study provides a baseline to guide LMICs globally in preservation of pediatric leukemias to preserve sample integrity for transportation time >48 hours. Future directions include evaluation of additional antibodies, testing these fixatives at higher temperatures, sample exchange with partner sites and method validation.

Presenter: Darong Yang, PhD

Mentor: Terrence L. Geiger, MD, PhD

Co-Authors: Song-Eun Lim, Yogesh Dhungana, Jeremy Chase Crawford, Jayadev Mavuluri, Pradeep Shukla, Lindsay L. Jones, Jiyang Yu, Terrence L. Geiger

Regnase-4 restricts CAR T cell longevity

Abstract
A lack of long-term persistence of chimeric antigen receptor (CAR) T cells is a major cause of relapses after CD19CAR T-cell therapy in B-cell acute lymphoblastic leukemia (ALL) patients. Our study has sought to identify new targets and combination strategies for enhancing CAR T cell persistence and function. The Regnase family includes four ribonucleases whose roles in T cells are unclear. In an immunocompetent murine ALL model, we demonstrate that the ablation of Reg4 (Regnase-4) in CAR T cells significantly increased ALL tumor control and prolonged mouse survival compared with non-edited CAR T cells. Reg4 knockout (KO) CAR T cells expanded 10-12 fold more than the control group in both spleen and bone marrow of tumor- bearing mice. Effector memory (TEM) and precursor exhaustion T cells (TPEX) were dramatically enriched in the Reg4-KO group. In a preclinical human CAR T model, CD19CAR T cells with Reg4 deletion markedly suppressed ALL tumor growth and improved NSG mouse survival. Double knockout (DKO) of Reg4 and Reg1 (Regnase-1) further enhanced CAR T-cell therapeutic efficacy compared with single manipulations. DKO CAR T cells persisted much longer and produced more type 1 and type 2 cytokines than control CAR T cells during repetitive antigen stimulations in vitro but after multiple rounds of stimulation only expressed high levels of type 2 cytokines. In patients treated with CD19CAR T cells on clinical trials, lower expression levels of Reg4 and Reg1 were associated with functional CAR T states and higher expressions correlated with terminal exhaustion features. Furthermore, long-lived CAR T cells in patients exhibited a type 2 immune phenotype and downregulated Reg4/1 expression. Our study demonstrates that Reg4 is a crucial negative regulator in T cell antitumor immunity. Targeting Reg4 alone or together with Reg1 is a promising approach to improve CAR T cell functionality and persistence and enhance its efficacy for cancer treatment.

Presenter: Alexa Siskar

Mentor: Jason Chiang, MD, PhD

Co-Authors: Alexa N. Siskar*, Emily Hanzlik*, Maria F. Cardenas, Mohammad K. Eldomery, Soniya Pinto, Christopher L. Tinkle, Qunyu Zhang, Xiaoyu Li, Tong Lin, Sandeep K. Dhanda, Gerald Reis, Daphne Li, Ravi Raghavan, Alexander Vortmeyer, Matthias A. Karajannis, Giles W. Robinson, Arzu Onar-Thomas, Patrick R. Blackburn, David A. Wheeler

FOXR2-activated CNS tumors exhibit substantial diversity across histological, molecular, and clinical dimensions

Abstract
FOXR2 activation is regarded as pathognomonic for CNS neuroblastoma (NB). However, a comprehensive understanding of the landscape for CNS tumors exhibiting FOXR2 activation is lacking.

Histopathologic, molecular, imaging, and clinical data of 42 FOXR2-overexpressing CNS tumors identified through screening of institutional datasets and published institutional cases were analyzed.

Among the 42 tumors, 21 (50.0%) were high-grade gliomas (HGGs), and 18 (42.9%) were embryonal tumors. The HGGs included ten H3 K27M-mutant diffuse midline gliomas (DMGs) and eight radiation-associated tumors. The embryonal tumors included 11 CNS NBs and six pineoblastomas (PBs). FOXR2 expression was similar between CNS NB and other tumor types (p = 0.82). FOXR2-overexpressing HGGs, unlike NBs and PBs, displayed diverse concomitant genetic alterations. The most common mechanisms of FOXR2 activation involved structural alterations causing promoter donation and enhancer hijacking from active genes essential for brain development, followed by alternative promoter activation or truncated LINE-1 retrotransposition. The preferential activation mechanism varied by tumor type. All but two aberrant FOXR2 transcripts incorporated non-canonical, non-coding exons. Gene set enrichment analysis demonstrated shared downstream effects of FOXR2 activation at the epigenome and transcriptome levels across tumor types. FOXR2-activated DMGs and PBs were aggressive, with 0% 2-year overall survival, whereas CNS NBs responded well to combined chemotherapy and radiation.

FOXR2-activated CNS tumors manifest significant histological, molecular, imaging, and clinical diversity. While FOXR2-activated HGGs and PBs demonstrated inferior prognosis, CNS NBs showed favorable outcomes. Integrating histologic and molecular diagnostic approaches is imperative for accurate prognostication and optimal therapeutic decision-making.

Presenter: Chun Shik Park, PhD

Mentor: Charles Mullighan, MBBS , MD

Co-Authors: Hao Shi, Ilaria Iacobucci, Qingsong Gao, Prady Baviskar, Zuo-Fei Yuan, Long Wu, Vishu Pagala, Kiran Kodali, Wei Wang, Andy High, Paolo Neviani, Yiming Wu, Lindsey Montefiori, Yunchao Chang, Tanya Khan, Hongbo Chi, and Charles Mullighan

The Impact of IKZF1 Alteration in the Tumor Microenvironment in BCR::ABL1 Acute Lymphoblastic Leukemia

Abstract
Alterations of IKZF1, encoding the lymphoid transcription factor IKAROS, are enriched in high-risk B-ALL (e.g. BCR::ABL1- positive and BCR::ABL1-like) and associated with poor outcome and drug resistance, including patients with BCR::ABL1 ALL treated with dasatanib and blinatumomab. We previously showed that IKZF1 alterations deregulate tumor cell adhesion and localization in the tumor microenvironment (TME). We hypothesized that IKZF1 alterations may deregulate the TME and thus blinatumomab responsiveness.

We generated a mouse model of the common IKZF1 deletion, IK6 (IKZF1flox[E4-E7]/+). Transplantation of bone marrow transduced with BCR::ABL1-Cre induced B-ALL with increased penetrance and reduced latency in immunocompetent recipients compared to wild type IKZF1. IK6 expression was associated with deregulation of the TME, including inhibition of neutrophil and B cell maturation, promotion of an anergic T cell phenotype, and upregulation of Treg cells and PD-L1+ macrophages. Deregulation of the TME was also observed in human BCR::ABL1;IK6 ALL, with reduced dendritic cells mediating tumor-derived antigen presentation to T cells, and impaired myeloid and phagocyte recruitment pathways. IK6 promoted immune evasion by inhibiting T cell activation and antigen presentation through downregulation of MHC-I in tumor cells, and CD8/NK cytotoxicity by downregulation of IFNγ.

Extracellular vesicles (EVs) are intercellular immune mediators by delivering protein and miRNA, and we observed that EVs from IK6 leukemic mice have elevated proteins involved in the regulation of cell proliferation. Using leukemic cells transduced with the EV reporter PalmGRET, we revealed that EVs from IK6 leukemic cells strongly interact with multiple immune cell types.

Collectively, IKZF1 alterations promote immune evasion in BCR::ABL1 ALL by blocking antigen presentation, maturation of neutrophil and B cells, and T cell activation, which may be mediated by inside-out signaling through EVs.

Ongoing work is examining blinatumomab responses according to IK1/6 status in vitro and in vivo in PDX and engineered mouse models; examining the requirement for EVs in TME remodeling using Rab27a-knockout (EV deficient) mice; and examining deregulation of the TME in human BCR::ABL1 ALL patient samples using spatial transcriptomics.

Presenter: Jonathan Sookdeo, MD

Mentor: Faizan Malik, MD

Co-Authors: Selene C. Koo, Faizan Malik

Expanding the Clinicopathologic Spectrum of CIC::NUTM1 Sarcoma in the Pediatric Population

Abstract
CIC::NUTM1-fused undifferentiated round cell sarcomas (URCS) are a subtype of CIC-rearranged URCS with few cases reported to date. We expand the clinicopathologic spectrum of this rare tumor in pediatric patients. 5 cases of molecularly confirmed CIC:NUTM1-fused URCS of bone/soft tissue were reviewed with the available imaging, histological, immunohistochemical, and clinical characteristics. The tumors occurred in 4 females and 1 male (age range of 1–14 years). Affected sites included paravertebral soft tissue (3 of 5), and deep soft tissue of the head/neck region (2 of 5). Two patients presented with immobility and hemiplegia, and one with bulging, painful mass. By imaging (available in 2 patients), the tumors involved the cervical-occipital paravertebral soft tissue and L4 paravertebral soft tissue in each. Histologically, all tumors were highly cellular with round (n=3), spindled (n=2), and rhabdoid (n=1) appearance of the tumor cells, in a sheeted (n=3) and trabecular (n=2) growth pattern. In 2 cases, the stroma had a myxoid to chondromyxoid appearance. Necrosis and high mitotic activity were present in all. Immunohistochemically, all tumors showed focal CD99 expression, with additional positivity for SMA (3/5), desmin (1/3), S100 (1/2), CD34 (1/2), WT1 (1/1), and NUT (4/4). The ad-hoc submitting diagnoses for molecular testing included mesenchymal chondrosarcoma, desmoplastic small round cell tumor, myofibroblastic sarcoma, and URCS. FISH testing confirmed CIC rearrangement in 2 cases; remaining 3 were confirmed by RNA sequencing. From our experience and literature review CIC::NUTM1 sarcomas represent an aggressive neoplasm with diverse morphologic features, typically presenting in a younger age group than CIC::DUX4 sarcomas, with a proclivity for the paravertebral and midline soft tissue sites. NUT immunopositivity represents a pitfall for misclassification as a NUT midline carcinoma.

Presenter: Katelyn S Provine, MD

Mentor: Faizan Malik, MD

Co-Authors: Selene C. Koo, Enrico Pozzo, Rita Alaggio, Faizan Malik

Abstract
Identification of fusions in infantile fibrosarcoma (IFS) and kinase-altered spindle cell neoplasms (SCNs) requires molecular assays that are either variable in performance, or not widely available. SRY-box transcription factor 11 (SOX11) is a transcription factor involved in the regulation of embryonic development. By transcriptomic analysis, we identified high levels of SOX11 expression in index IFS cases with ETV6::NTRK3 fusions. This prompted us to investigate SOX11 expression pattern by IHC across a cohort of IFSs, NTRK-rearranged SCNs, and other morphologic mimics to assess its predictive utility. SOX-11 IHC (MRQ- 58, RTU, Cell Marque) was performed on 124 cases. These included whole tissue sections (WTS) of 37 tumors including pediatric ETV6::NTRK3-fused IFS/SCNs (n=19), adult and pediatric NTRK1-rearranged SCNs (n=12), SCNs with non-ETV6::NTRK3 fusions (n=2), inflammatory myofibroblastic tumors (n=4) (IMTs), and tissue microarrays (TMAs) of 87 histologic mimics, including 19 desmoid fibromatoses (n=19), schwannomas (n=4), malignant peripheral nerve sheath tumors (n=33), neurofibromas (n=7), and synovial sarcomas (n=24). Positive staining threshold was set to >10% of neoplastic cells at moderate staining intensity in the nuclei. SOX11 IHC was negative in all 12 NTRK1-rearranged SCNs, both EML4::NTRK3 and RBPMS::NTRK3-fused SCNs, and all 4 IMTs. SOX11 IHC was positive in 63% (12/19) of ETV6::NTRK3-rearranged SCNs, staining diffusely (>50% of neoplastic nuclei) in 42% (5/12) and focally (10-50% of neoplastic nuclei) in 58% (7/12). SOX11 was also positive in 21% (5/24) of synovial sarcomas (total 5/87 positive tumors in the TMA cohort). Given the expression pattern, SOX11 appears to be a highly specific but moderately sensitive marker for ETV6::NTRK3 rearranged SCNs. In the correct morphologic context, SOX11 shows utility as a screening adjunct for focused molecular testing, especially in resource-limited settings.

Presenter: Patrick R. Blackburn, PhD

Mentor: Faizan Malik, MD

Co-Authors: Zonggao Shi, Mohammad K. Eldomery

Clinical and molecular insights into USP6-associated neoplasms with novel or rare fusion partners in a pediatric cohort

Abstract
USP6-associated neoplasms constitute a spectrum of related entities that share clinical and morphological characteristics and are defined molecularly by the presence of a USP6 gene rearrangement. These rearrangements usually result in a promoter swap fusion involving a range of partner genes that activate ectopic expression of full-length USP6. Initially, described in aneurysmal bone cyst (ABC), the spectrum of USP6-associated neoplasms has expanded to include nodular fasciitis, cranial fasciitis, myositis ossificans, and fibro-osseous pseudotumor of digits, and cellular fibroma of the tendon sheath. USP6 rearrangements have also been described in a subset of aggressive myofibroblastic neoplasms. In this study, we retrospectively reviewed whole transcriptome results in the St. Jude clinical cohort and identified a series of USP6-associated neoplasms with novel or rare fusion partners. Novel USP6 fusion partners included SEC24D, HNRNPC, and ERRFI1 identified in three cases of ABC, including one case with prominent myxoid features. We describe the clinical, radiographic, and histologic presentations of these new cases and describe the associated fusion structures and provide a brief review of the literature.

Presenter: Ifenna Amaefuna

Mentor: Gabriela Gheorghe, MD 

Co-Authors: Heather Shepard, Mahsa Khanlari, Ruth Wang’ondu, Mathew J. Ehrhardt

B7-H3 and CD70 expression in pediatric ALK+ anaplastic large cell lymphoma: A single institutional experience 

Abstract
B7-H3 (PD-L3) and CD70 are immune checkpoints whose expression has recently been documented in a variety of malignant neoplasms and is associated with a worse prognosis for patients. Although several studies have shown the efficacy of immunotherapies that target these surface antigens, there is a paucity of literature describing B7-H3 and CD70 expression in pediatric anaplastic lymphoma kinase (ALK) + anaplastic large cell lymphoma (ALCL).  By evaluating the expression of these antigens in ALCL, we hope to contribute towards evolving literature and encourage the development of new targeted therapies. 

We retrospectively examined the expression of B7-H3 and CD70 in immunohistochemically stained tissue samples from 31 pediatric patients with de novo and/or relapsed anaplastic lymphoma kinase (ALK)-positive anaplastic large cell lymphoma (ALCL). B7-H3 and CD70 expression was scored using intensity (negative, +1, +2, and +3), % of expression, and H-score in neoplastic cells. 

The immunostain for CD70 was taken up primarily by cell membranes and Golgi bodies while B7-H3 was localized to cell membranes. Initial immunohistochemistry (IHC) analyses by a board-certified hematopathologist identified 13 (41.9%) and 10 (32.3%) patients showcasing B7H3 and CD70 positivity respectively. The median expression of CD70 was 40% and the median H-score was 110.

Our results show moderate expression of CD70 and B7-H3 in ALK positive ALCL cells. This underscores the significance of studying these antigens in pediatric ALK+ ALCL for the future design of targeted therapies. Further studies are needed in large pediatric cohorts to confirm our findings. However, our data supports the use of B7-H3 CAR-T cells in ALK + ALCL. 

Presenter: Julieann C. Lee MD

Co-Authors: Julieann C. Lee, Quynh T. Tran, Rose B. McGee, Melissa R. Perrino, Santhosh A. Upadhyaya, Emily M. Hanzlik, Nicholas Pytel, Andrew J. Carroll, Wilda Orisme, Mohammad Eldomery, Lu Wang, Patrick R. Blackburn, Larissa V. Furtado, Angela N. Viaene, Minjie Luo, Jennifer M. Kalish, Soniya N. Pinto, Asim K. Bag, and Brent A. Orr

Atypical teratoid/rhabdoid tumor-TYR subtype arising in the setting of germline ring chromosome 22: an uncommon form of tumor predisposition

Abstract
Atypical teratoid/rhabdoid tumor (ATRT) is an embryonal neoplasm that often contains rhabdoid cells and polyphenotypic differentiation. Biallelic inactivation of SMARCB1, or less commonly SMARCA4, is the defining molecular alteration of ATRT. While rhabdoid tumor predisposition syndrome with germline inactivation of SMARCB1 or SMARCA4 is widely recognized and evaluated for in patients with ATRT, a less common form of ATRT predisposition is the consequence of germline ring chromosome 22. Germline ring chromosome 22, though not disruptive of SMARCB1 directly, is inherently unstable and prone to loss creating a predisposition to somatic mosaicism for monosomy chromosome 22. Deletion or disruption of distal genes such as SHANK3 during ring formation of chromosome 22 can result in Phelan McDermid syndrome (OMIM #606232) or other phenotypic sequelae. To raise awareness of this alternative mechanism for germline predisposition to ATRT and highlight associated clinical manifestations, we present a case with detailed tumor and germline molecular characterization.

Presenter: Julieann C. Lee MD

Co-Authors: Julieann C. Lee, Amar Gajjar, Melissa R. Perrino, Angela Delaney, Christopher L. Tinkle, Paul Klimo, Larissa V. Furtado, Liu, Yen-Chun MD, Alex Breuer, Soniya N. Pinto, Asim K. Bag, David W. Ellison, and Brent A. Orr

Concurrent Pit-1 and SALL4 staining in primitive component of pituitary blastoma: a case report with clinicopathologic features including Cushing syndrome presentation, germline DICER1 mutation, and significant therapeutic response to chemoradiation

Abstract
Pituitary blastoma is a rare CNS entity of the sellar region, initially described by Dr. Scheithauer in 2008, with reported cases in the literature currently approaching 20. The three histologic features of pituitary blastoma are 1) a primitive embryonal component, 2) a neuroendocrine component, and 3) a Rathke-type epithelium. Typically occurring in children less than 2-years of age, with Cushing syndrome being a common clinical presentation, pituitary blastoma is known to occur in association with germline DICER1 mutation. Given the rarity of this entity, the clinical, histologic, and imaging features are shared here for an additional patient with pituitary blastoma, including the novel immunohistochemical finding of concurrent SALL4 and Pit-1 immunostaining within the primitive component of the tumor. Prior reports of pituitary blastoma with evaluation of germ cell tumor markers such as PLAP and others, did not include SALL4 staining results. As SALL4 can be seen to some degree in a variety of embryonal tumors, our findings suggest SALL4 expression may serve as a useful diagnostic method to establish a primitive phenotype in ambiguous cases. Continued reporting of clinical treatment and outcomes for pituitary blastoma patients will add to prognostic and management knowledge for this rare entity, which in this case after subtotal resection showed significant therapeutic response to induction chemotherapy and focal proton radiation.

Presenter: Vidya Balagopal PhD

Mentor: Mohammad K. Eldomery, MD

Co-Authors: Debbie Payne-Turner, Alexander Breuer, Jamie Maciaszek, Lu Wang, Jeffery M. Klco, Swati Naik , Aimee Talleur, Paulina Velasquez

Development and Validation of a Multiplexed ddPCR Assay for Monitoring of CAR-T Cell Persistence in Clinical Settings

Abstract
Chimeric Antigen Receptor T-cell (CAR-T) therapy utilizes genetically engineered T-cells to target and destroy cancer cells. The success of CAR-T treatment relies heavily on the engraftment, expansion, and persistence of CAR-T cells within the patient. Thus, implementing precise and quantitative monitoring of CAR-T cells post-infusion in a clinical setting is essential. Currently there is no clinically validated assay for CAR-T cell enumeration at St. Jude Children’s Research Hospital, although a quantitative PCR-based method specific to each CAR construct is available as a research assay.

Digital droplet PCR (ddPCR) is a robust, sensitive assay that enables absolute quantification of nucleic acids by partitioning the sample into thousands of individual droplets. Each droplet undergoes PCR amplification independently, allowing for the detection of very low-abundance target molecules without the need for standard curves. We have validated, for clinical use, a multiplexed quantitative ddPCR assay on the Bio-Rad digital PCR low-partitioning platform to monitor lentiviral-based CAR-T cell therapies. By targeting Psi (Ψ) as a universal marker, we can detect any CAR-T cell therapy that uses a lentiviral vector backbone. Using just 300ng of total DNA input, we are able to detect as low as 0.01% consistently. From studying a cohort of over 40 samples, we show excellent specificity, sensitivity, and reproducibility of CAR-T cell detection. The assay is currently validated for bone marrow and peripheral blood samples.

In conclusion, this multiplexed ddPCR assay is a robust and reliable tool for clinically monitoring CAR-T cell persistence in blood and bone marrow, supporting early and long-term follow-up in clinical trials. Future directions include expanding to CSF samples and to quantifying CAR-T cells with respect to overall T-cells, allowing precise measurement of the proportion of genetically modified T-cells.

Presenter: Mohammad K. Eldomery, MD

Co-Authors: Jamie L Maciaszek, Taylor Cain, Victor Pastor Loyola, Suraj Sarvode Mothi, David A Wheeler, Li Tang, Lu Wang, Jeffery M Klco, Patrick R Blackburn

Evaluation of Bayesian point-based system on the variant classification of hereditary cancer predisposition genes

Abstract
Purpose: To assess the differences in variant classifications using the American College of Medical Genetics and Genomics and the Association for Molecular Pathology 2015 guidelines and the Bayesian point-based classification system (here referred to as the point system) in 115 hereditary cancer predisposition genes and explore variant sub-tiering by the point system.

Methods: Germline variant classifications for 721 pediatric patients from an in-house panel were retrospectively evaluated using the 2 scoring systems.

Results: A total of 2376 unique variants were identified, with ∼23.5% exhibiting discordant classifications. Unique variants classified by the point system demonstrated a lower rate of variants of uncertain significance (VUS;∼15%) compared with American College of Medical Genetics and Genomics and the Association for Molecular Pathology 2015 guidelines (∼36%). This change is attributed to unique variants with 1 benign supporting evidence (∼12%) or 1 benign strong evidence (∼4%) being classified as likely benign by the point system. Additionally, variants with conflicting/modified evidence (∼5% of 2376) are also resolved by the point system. Sub-tiering unique variants classified by the point system as VUS (n = 354) indicates ∼77.4% were VUS-Low (0-1 points), whereas the remaining ∼22.6% were VUS-Mid (2-3 points) and VUS-High (4-5 points).

Conclusion: The point system reduces the VUS rate and facilitates their sub-tiering. Future large-scale studies are warranted to explore the impact of the point system on improving VUS reporting and/or VUS clinical management.

Presenter: Mehul Jani, PhD

Mentor: Selene Koo, MD, PhD

Co-Authors: Liyuan Li, Vidya Balagopal, Jamie Maciaszek, Ruth Tatevossian, Lu Wang, Jeffery Klco

Development of a clinical assay for low-frequency variant detection in circulating tumor cell-free DNA in pediatric solid tumors

Abstract
Analysis of tumor-derived circulating cell-free DNA (ctDNA) is increasingly being used to monitor minimal residual disease (MRD) of tumor and detect clinically relevant novel variants in recurrent/residual or metastatic tumors. In pediatric cancer, research approaches have been developed to detect selected alterations, including targeted single nucleotide variants (SNVs), few gene fusions, and copy number changes in limited tumor types like neuroblastoma and rhabdomyosarcoma. We sought to develop and clinically validate a comprehensive assay for the detection of single nucleotide variants and small insertion-deletions (SNVs/Indels), gene fusions, and copy number alterations (CNA) relevant for a wide variety of pediatric solid tumors to monitor MRD and tumor evolution.

Our methodology allows very low input amount of DNA and uses dual unique molecular identifiers to reduce sequencing errors. We combine hybrid capture using the SJPedPanel and low pass whole genome sequencing to detect different variant types of clinical interest, including SNVs/Indels, gene fusions and CNAs relevant across a broad spectrum of pediatric solid tumors. The bioinformatics pipeline integrates the DRAGEN platform for SNV and Indel detection, cfdna_cnv for CNV detection, and Aperture for precise structural variants detection. We have developed a workflow for clinical validation sample collection from patients with different tumor types. On our test cohort of patient samples and reference material, we demonstrate our method’s ability to detect low-frequency variants down to 0.5% allele frequency. We also test clinically relevant parameters, including accuracy and precision, for the detection of variants matched to those seen in tumor tissue as well as clinically relevant de novo mutations.

This study highlights the effectiveness of liquid biopsy and advanced bioinformatics for detecting low-frequency variants in solid tumor ctDNA, offering a promising tool for precision oncology at St. Jude.

Presenter: Rebecca Voss, MD

Mentor: Lu Wang, MD, PhD

Co-Authors: Meiling Jin, Yen-Chun Liu, Shaohua Lei, Caleb Gallops, Michael Rusch, Mark R. Wilkinson, David Rosenfeld, Priya Kumar, JeRery M. Klco, Charles G. Mullighan, Lu Wang

Enhancing Molecular Diagnostic Workup: A Novel Method for Comprehensive WGS-based DNA Index and Ploidy Analysis with a 5-Day Turnaround Time

Abstract
Determining tumor ploidy plays an important role in the diagnostic workup and risk stratification of certain types of cancer, such as B-lymphoblastic lymphoma (B-ALL). Commonly, DNA index by flow cytometry (Flow-DI) is used to determine ploidy; yet, this method has limitations as it can misdiagnose masked hypodiploidy/near-haploidy as hyperdiploidy. Here, we present a novel method using whole genome sequencing to measure DI (WGS-DI), providing detailed information on ploidy changes, and introduce a workflow in which results can be achieved with a turnaround time of 5 working days.

Genomic profiling by WGS and Flow-DI were performed at diagnosis for tumor samples from 58 pediatric B-ALL patients with various ploidy changes. WGS data was processed by a standard mapping pipeline (≥24h) and additionally, for a subset of 16 patients via the accelerated Illumina DRAGEN mapping pipeline (2-3h). DNA copy number analysis was conducted via the DRAGEN pipeline, followed by internal analysis to identify copy number variants (CNV) and to generate CNV/B-allele frequency plots. DNA content was determined by adjusting the pre-established, gender-matched reference DNA content to DNA gains or losses detected by WGS. WGS-DI was calculated as the ratio of test sample DNA content to reference DNA content.

The obtained WGS-DI values closely aligned with Flow-DI values in 53 of 58 cases (r=0.99). The remaining 5 cases, identified as masked low-hypodiploid, showed WGS-DI below 0.75, reflecting the original clone before genome doubling. Review of CNV and B-allele frequency plots identified 2 more masked low-hypodiploid cases, with WGS-DI above 1.1 reflecting actual tumor DNA content. The comparison between fast and standard mapping approaches showed nearly identical DI values, demonstrating the feasibility of using WGS and fast mapping as an alternative for DI calculation.

WGS-based DNA indexing reliably identifies hyperdiploid B-ALL cases and accurately distinguishes masked low-hypodiploid/near-haploid cases from hyperdiploidy. Our pilot study demonstrates that accelerated Illumina DRAGEN WGS mapping can provide CNV data consistent with results derived from standard mapping and, that the total turnaround time of the new WGS-DI workflow is comparable to the standard Flow-DI with improved accuracy of ploidy interpretation.

Presenter: Zahangir Alom, PhD

Mentor: Brent A. Orr, MD, PhD

Co-Authors: Quynh T. Tran and Brent A. Orr 

Pathology Stable Diffusion Model for Synthetic Histology Image Generation

Abstract
Background: Computational pathology using Deep Learning (DL) has potential to improve histopathologic diagnosis by augmenting critical activities related to diagnosis, prognostication, and prediction of therapeutic response. One of the main challenges to implementation of DL-based systems for clinical applications is acquiring and annotating sufficient training examples. Whole Slide Images (WSI) are significantly larger (~400x) than other medical imaging modalities, requiring patch sampling during the pre-processing phase. As a result, sample preparation and annotation are complex, time-consuming, and costly for development of pathology applications. In addition, the rarity of pediatric cancers makes acquisition of sufficient training examples to develop AI systems difficult. To address these issues, we propose a stable diffusion-based pathology image generation method, to generate plausible synthetic histology images that might be used to train DL classification models.

Methods: We explored the ability of stable diffusion models to generate plausible histology patch images following training on three reference datasets, including colon cancer (262,777 examples), multiple brain tumor types (120K examples), and medulloblastoma molecular subtypes (56K examples). The synthetic image quality was assessed qualitatively by a board-certified pathologist. The image quality was quantified by comparing the classification scores of an independent supervised classification model and the deep learned feature embedding of the synthetic versus reference data.

Results: We successfully generated plausible histology images using a stable-diffusion-based model. The model demonstrated good ability to generate colon cancer images but showed more variability in generating specific brain tumor types. While the deep learned feature representations of synthetic data showed significant overlap with reference data, supervised models frequently misclassified synthetic images compared to reference data.

Conclusion: We demonstrate the feasibility of using generative models to create synthetic histopathology images. Future efforts will improve on the fidelity of synthetic images by using more highly curated reference data and employing next generation diffusion models.

Presenter: Quynh T. Tran, PhD

Mentor: Brent A. Orr, MD, PhD

Co-Authors: Md Zahangir Alom and Brent A. Orr

Methylation of the TERT promoter and the risk stratification in uveal melanoma

Abstract
Background: Uveal melanoma (UM) is the most common and aggressive intraocular malignancy in adults, with two prognostic groups based on gene expression profiles: class 1 (good prognosis) and class 2 (poor prognosis). The poor outcome in Class 2 UM is associated with their high metastatic potential. While telomere maintenance has been linked to metastasis in other tumor types, the mechanism of telomere regulation in UM remains unclear. Methylation of the telomerase reverse transcriptase (TERT) hypermethylated oncological region (THOR) is a frequent gain-of-function mechanism in cancer. We hypothesized that THOR hypermethylation regulates telomeres in UM and mediates aggressive behavior in high-risk disease.

Methods: Unsupervised analysis of 80 uveal melanomas from The Cancer Genome Atlas identified two methylation groups (high- and low-risk). Mutation data in TERT and ATRX were evaluated. THOR methylation levels were assessed using probe methylation levels at cg11625005, and TERT expression was evaluated relative to DNA methylation levels. We analyzed relationships between survival, methylation class, and THOR levels, and tested the predictive ability of these markers in an independent validation cohort.

Results: High-risk patients showed 16.9 times higher death hazard than low-risk patients (p=0.000143). No TERT or ATRX were observed in the TCGA cohort. THOR methylation levels correlated positively with TERT expression, and high THOR levels predicted poor outcome. Within the high-risk group, THOR hypermethylation increased death hazard 5.47-fold compared to hypomethylation (p=0.00216). In an independent validation dataset, methylation class significantly predicted metastatic disease, while THOR hypermethylation was not an independent predictor when accounting for methylation class.

Conclusions: Methylation class predicts outcome and metastatic propensity in UM, while THOR hypermethylation influences survival in high-risk class 2, it is not an independent predictor of metastatic disease.

Presenter: Jason Chiang, MD, PhD

Co-Authors: Dana Tlais*, Jordan T. Roach*, Christopher L. Tinkle, Qunyu Zhang, Xiaoyu Li, Tong Lin, Ayatullah Mostafa, Daniel C. Moreira, Rene Y. McNall-Knapp, Sarah Z. Rush, Brian H. Le, Sara Sinno, Apeksha Agarwal, Kevin F. Ginn, Richard M. Green, Sonia Partap, Arzu Onar-Thomas, Asim K. Bag

Clinical Features, Prognostic Factors, and Pattern of Failure in H3 G34-Mutant Diffuse Hemispheric Glioma: A Multi-Institutional Experience and Meta-Analysis

Abstract
Background: H3 G34-mutant diffuse hemispheric glioma (DHG) is an aggressive tumor with a poor prognosis. We investigated the relationship between patient outcomes and molecular characteristics, the extent of resection, and temozolomide (TMZ) use. Additionally, we reviewed the patterns of treatment failure.

Methods: Retrospective multi-institutional review of clinical, imaging, and molecular characteristics of 38 patients with newly diagnosed H3 G34-mutant DHG, supplemented by a meta-analysis.

Results: The median age was 14 years (8-28 years). The median progression-free survival (PFS) was 0.6 years (95% CI 0.4-1.2 years), and the median overall survival (OS) was 1.8 years (95% CI 1.1-3.0 years). Gross total resection (GTR) was associated with improved PFS (p=0.0078) compared to non-GTR. Twenty-two patients (57.9%) received frontline TMZ and had improved PFS compared to those without (p=0.0034). Of the evaluable patients with progressive disease, 78% progressed within the high-dose RT field. MGMT promoter methylation was not significantly associated with PFS/OS or TMZ efficacy (33 evaluable cases, n=19 with MGMT silencing, n=14 without). PDGFRA amplification (n=10) was associated with inferior OS (p=0.0443), and CDKN2A homozygous deletion (n=16) was associated with inferior PFS (p=0.0204). In the meta-analysis cohort (n=252), GTR/near- total resection had significantly better PFS (p<0.0001) and OS (p<0.0001), and MGMT promoter methylation was not associated with PFS/OS.

Conclusions: In our cohort, MGMT promoter methylation was not a prognostic factor and was not associated with TMZ utility. The resection extent and TMZ use were associated with improved survival outcomes. As most treatment failures occurred within the high-dose RT field, extended fields are not warranted.

Presenter: Brent A. Orr MD, PhD

Co-Authors: Md Zahangir Alom, Quynh T. Tran

Scalable large-scale supervised histology image classification using foundational model representation learning

Abstract
Background: Deep learning approaches for histopathological classification of brain tumors have shown promise, but most models are restricted to only a few tumor classes, show significant variability in generalizability across unseen datasets, and struggle to scale to handle the breadth of tumor types encountered in clinical practice. Several foundational models have been developed that help address issues related to scalability and generalizability. Foundational models, which are self-supervised models trained on a high number of histology slides, can be used to extract reproducible learned histology image representations for downstream supervised or unsupervised tasks. In this study we compared the performance of large-scale tumor classification models, consisting of 74 distinct brain tumor classes, trained on the deep learned feature representations from multiple self-supervised foundational models.

Methods: We employed four pre-trained self-supervised models (RetCLL, UNI, Prov-GigaPath, and DeepCMorph) to extract deep feature representations from histological images. These features were then used to train supervised deep neural networks for tumor classification, with model performance assessed using accuracy, precision, recall, and F1-score on hold-out test data. Comparisons of the model convergence rate and the learned representation embeddings were performed to evaluate the relationship of these features to model performance.

Results: The UNI and Prov-GigaPath models demonstrated superior performance, achieving test accuracies of 92.07% and 93.31%, respectively, with corresponding F1-scores of 0.921 and 0.933. RetCLL showed moderate performance (82.38% accuracy, 0.824 F1-score), while DeepCMorph achieved the lowest performance characteristics (79.40% accuracy, 0.795 F1-score). Evaluation of model convergence rates and the unsupervised image embeddings suggest the quality of the learned representations correlates closely with the supervised model performance.

Conclusions: Our results suggest that representation learning from self-supervised foundational models is a feasible and effective approach to train large-scale brain tumor classification models. Future work will focus on validating generalizability across independent datasets.

Presenter: Jayadev Mavuluri, PhD,

Mentor: Terrence Geiger, MD, PhD

Co-Authors: Raj Alli; Darong Yang

Empowering CD19-CAR-T Therapy: Precision Mutation on CD19 in Primary B-Cells to Prevent Prolonged B-Cell Aplasia in B-ALL Patients

Abstract
CD19-directed Chimeric Antigen Receptor T-cell (CAR-T) therapies, particularly FMC63-based CAR constructs, have demonstrated significant efficacy in treating B-cell malignancies. However, a major side effect is B-cell aplasia, resulting from the depletion of normal CD19-expressing B cells, leading to hypogammaglobulinemia, increased infection risks, and a need for immunoglobulin replacement therapy. This study addresses B-cell aplasia by engineering the CD19 gene in B cells to reduce CAR-T recognition while preserving essential CD19 functions. Functional assays confirmed that these engineered B cells retain normal immune activity and resist CD19-CAR-T cell cytotoxicity both in vitro and in vivo.

We extended this approach using CRISPR-Cas9 and prime editing on hematopoietic stem cells (HSCs), removing the FMC63-CAR binding motif on CD19 while allowing full B-cell functionality. These edited HSCs were successfully engrafted in immunodeficient mouse models, producing functional B cells, restoring immunoglobulin levels, and demonstrating minimal off-target effects. This approach, using CD19 locus-edited HSCs, offers a potentially translatable solution for sustaining CAR-T therapy benefits by preventing prolonged B-cell aplasia, supporting long-term immune recovery, and enhancing patient quality of life.

Presenter: Masayuki Umeda, MD, PhD

Mentor: Jeffery M. Klco, MD, PhD

Co-Authors: Ryan Hiltenbrand, Nicole Michmerhuizen, Juan Barajas, Michael P Walsh, Guangchun Song, Jing Ma, Tamara Westover, Ilaria Iacobucci, Charles G. Mullighan, Jeffery M. Klco

Cooperating alterations determine cellular hierarchies of NUP98::KDM5A models

Abstract
Fusion partners of NUP98-rearranged (NUP98r) acute myeloid leukemia (AML) are associated with various disease types and cooperating mutations, such as RB1 loss in NUP98::KDM5A acute megakaryocytic leukemia (AMKL). Despite these associations, how fusion partners and cooperating mutations contribute to the disease phenotypes remains unclear. To study the effect of cooperating alterations, we first established cord blood CD34+ (cbCD34) NUP98::KDM5A and NUP98::NSD1 models with Cas9 expression, followed by transduction of gRNA targeting WT1 and RB1 so that they mimic frequent cooperating alterations (RB1 loss and WT1 exon 7 frameshift). WT1-targeting gRNAs lead to growth advantages over AAVS-targeting gRNA in both NUP98::KDM5A and NUP98::NSD1 cbCD34 models with a higher ratio of cells with a CD34+CD41a- immunophenotype. Contrarily, RB1-targeting gRNA induced enrichment of out-of-frame insertions and deletions (indels) only in the cbCD34 NUP98::KDM5A model accompanied by increases in a CD34+CD41a+ population, suggesting fusion partner-specific function of cooperating alterations. scRNAseq from the AAVS-targeting condition collectively showed differentiation trajectories resembling normal hematopoiesis, including an enrichment at the differentiated platelet cluster characterized by high RB1 and low CDKN2A expression. In contrast, RB1-gRNA conditions enriched cells at the MEP and megakaryocytic progenitor stages and lacked the platelet cluster, suggesting a differentiation block associated with RB1 loss. WT1-gRNA conditions showed enrichment of GMP-LMPP clusters at the expense of erythroid/megakaryocytic lineages, collectively suggesting the differentiation stage-specific effect of cooperating alterations in cbCD34 NUP98 models. Treatment of these cells with a menin inhibitor (revumenib) revealed that cbCD34 NUP98::KDM5A models with RB1- or WT1-gRNA are less sensitive to menin inhibition with increased CD34+CD41a+ population in the RB1-gRNA conditions. Our findings illustrate how cooperating mutations in NUP98r AML distinctly alter cellular differentiation hierarchies, with potential implications for menin inhibitor resistance mechanisms associated with specific differentiation stages and mutational profiles.

Presenter: Katelyn Purvis, MD

Mentor: Jeffery M. Klco, MD, PhD

Co-Authors: Tami Westover, Guanchung Song, Mike Walsh, Xiaotu Ma, Jing Ma, Pandurang Kolekar, Thomas Alexander, Norman Lacayo, John Choi, Seth Karol, Chin-Hon Pui, Jeffrey Rubnitz, Raul Ribeiro, Jeffery Klco

Molecular Response of Pediatric AML to Selinexor: Results of the SELHEM Trial

Abstract
With contemporary therapy, patients with pediatric acute myeloid leukemia (AML) experience survival rates of around 70%, with relapse rates between 20% to over 30%. Long-term survivors among relapsed patients are rare, with 5-year overall survival of only 35%. There’s a pressing need for broadly applicable targeted agents to enhance survival in this challenging disease. Selinexor (KPT-330), a selective exportin 1 (XPO1) inhibitor, shows promise. The SELHEM trial, a Phase I/II study at St. Jude Children’s Research Hospital, evaluated selinexor in combination with fludarabine and cytarabine in patients with relapsed/refractory leukemia (N=54). Therapy began with a 15-day selinexor-only phase, followed by combination therapy with fludarabine and cytarabine. Samples were collected before therapy, on day 15, and at treatment's end to evaluate clinical responses. Using genomic information from pre-enrollment whole genome sequencing, we designed a custom targeted panel comprised of 1-7 (mean=4) pathogenic/likely pathogenic variants and targeted resequencing (median depth~800x) was performed on day 15 (n=34 samples) and end-of therapy samples (n=30 samples) to monitor changes in variant allele frequencies (VAF) as a molecular biomarker of response. Patients with sequencing available were included in the genomics portion of the study (n=31). Sixteen patients (51.6%) exhibited molecular response (MR) defined as a 1,000-fold decrease in VAF at the end of therapy. Among these, 5 patients (31.3%) demonstrated early molecular response (EMR) to selinexor alone defined as a 3-fold decrease in VAF at day 15. Two patients achieved EMR without reaching MR at the end of therapy. Two patients with MR did not have day 15 samples available. Identified leukemic drivers among those with molecular responses to selinexor alone included NUP98::ZFX, RUNX1::RUNX1T1 (2 patients), CBFB::MYH11 (2 patients), KMT2A::MLLT3, and KMT2A::MLLT1. Acquired XPO1 mutations were not seen among non-responders. Single cell RNA sequencing for paired responders and non-responders is ongoing to further define molecular biomarkers of response to selinexor.

Presenter: Lisett Contreras, PhD

Mentor: Jeffery M. Klco, MD, PhD

Co-Authors: Michael P Walsh; Juan Martin Barajas; Tamara Westover; Melvin Thomas III

Investigating the in vivo impact of UBTF-TD in leukemogenesis and hematopoietic development

Abstract
UBTF-TD AMLs are recognized as a clinically relevant entity of pediatric AML. The presence of UBTF-TD in patients is an independent predictor of poor response to therapy and overall inferior outcome. Functional in vitro studies have been accomplished to investigate the role of UBTF-TDs in leukemogenesis. However, robust in vivo studies have yet to be established.

UBTF-TDs are defined by somatic tandem duplications (TDs) within exon 13 of UBTF (upstream binding transcription factor). UBTF-TD alterations are commonly found with FLT3-ITD and WT1 mutations but with no driver mutations. Our previous data suggest that these alterations are driver mutations for a new AML subtype associated with poor clinical outcomes.

The mechanism by which these tandem duplications alter hematopoietic development and drive leukemia development under physiological conditions is yet to be evaluated. In these studies, we utilize and characterize a newly generated conditional knock-in mouse model, Ubtf-cKI-TD84, which expresses a UBTF tandem duplication in the hematopoietic compartment from the endogenous Ubtf locus following exposure to Vav1- Cre. We investigate the impact of Ubtf-TD84 (from Ubtf-cKI-TD84 x Vav1-Cre mouse) on hematopoietic development through multiparameter flow cytometry and RNA sequencing.

We observed that hematopoietic cells from the Ubtf-TD84 mouse have a myeloid lineage bias. Ubtf-TD84 mice develop a myeloid leukemia without maturation as early as 20 weeks with variable penetrance. Tumor cells from these mice serially transplant and display dysregulated HOX gene cluster expression. Further analyses will be accomplished to examine the effects of introducing cooperating mutations and to identify potential vulnerabilities of UBTF-TD pathogenesis.

Presenter: Peixin Liu, PhD, SCYM

Mentor: Paul E. Mead, PhD, SCYM

Co-Authors: Laura Key, Christy Embrey, Stacie Woolard, Mahsa Khanlari, Priyadarshini Kumar, Jeffery M. Klco, Paul E. Mead

Development of a high-parameter spectral panel for determining Minimal Residual Disease (MRD) in pediatric T-acute lymphoblastic leukemias (T-ALL)

Abstract
Introduction: Flow cytometric assessment of residual disease during chemotherapy provides a robust indicator of treatment efficacy and risk of relapse in pediatric leukemia patients. In our continuing efforts to improve the clinical immunopathology laboratory workflow, we are developing high-parameter spectral panels to replace multi-tube 8- colour conventional flow cytometry methods. Here, we present a single-tube, thirty-nine parameter panel for the measurement of minimal residual disease (MRD) in pediatric patients being treated for T-cell acute lymphoblastic leukemia (T-ALL).

Methods: The single-tube spectral T-ALL MRD panel has four intra-cellular markers (TdT, cyCD79a, cyCD3, and MPO), thirty-three surface markers (CD1a, CD2, sCD3, CD4, CD5, CD7, CD8, CD10, CD11c, CD13, CD14, CD15, CD16, CD19, CD20, CD33, CD34, CD38, CD45, CD48, CD52, CD56, CD94, CD99, CD117, CD123, CD133, CD235a, CD335, HLA-DR, TCR-, TCR-, TCR CB1), a viability (Zombie NIR) and a nuclear marker (Draq5).

Results: Iterations of the T-ALL MRD spectral panel were tested with peripheral blood (PB) and bone marrow (BM) samples from T-ALL patients at early treatment stages. The inclusion of viability and nuclear markers allow accurate gating on the live-nucleated cellular component of the sample and provides a direct assessment of the denominator for the MRD calculation. Spectral cytometry provides the added ability to identify, and gate out, multiple auto-fluorescent populations that may interfere with determination of the blast count. We compared MRD values calculated using the conventional flow cytometry assay with the new spectral panel and showed excellent concordance between the two methods.

Conclusions: Spectral flow cytometry is ideally suited to MRD analysis. The ability to include many more markers in a single tube allows for accurate determination of blast cell numbers in the context of post-chemotherapy changes in normal populations. For example, we have included additional markers for NK cells that can confound blast estimates at later treatment timepoints in T-ALL. Our goal is to migrate all our multi-tube conventional flow cytometry panels to the high-content spectral platform. This will provide many benefits to our laboratory workflow and to the accuracy of downstream analysis.

Presenter: Zhaohong Yin, MD

Mentor: Paul E. Mead, PhD, SCYM

Co-Authors: Clair M. Kelley, Cyrus (Mo) Mehr, Priyadarshini Kumar, Jeffery M. Klco, Paul E. Mead

An improved method to preserve Cerebral Spinal Fluid (CSF) sample integrity for flow cytometric determination of blast counts and immunophenotype

Abstract
Introduction: Flow cytometric assessment of cerebral spinal fluid (CSF) samples in pediatric leukemia patients provides valuable information to the clinical team for the management of these patients. CSF flow is difficult due to the generally low cellularity of sample and the rapid degradation of sample quality. Fresh CSF samples must be assayed within hours of collection to avoid loss of cellularity and to maintain antigen stability. We investigated the use of two commercially available fixatives that could be used at the time of collection to improve sample stability and prolong the window for flow cytometry.

Methods: Flow cytometry assay development for CSF is hampered by the paucity of cells and the small volumes of individual leftover samples available for analysis. To circumvent these limitations, we pooled leftover ‘zero count’ CSF samples to attain sufficient volume to test multiple conditions and spiked in a known number of blasts, either 100 or 250 blasts per mL, using fresh diagnostic pediatric B-ALL or AML samples. Aliquots of these spiked samples were either left untreated or fixed with Transfix or Streck preservatives and stored for several days according to the manufacturer’s instructions. Flow cytometry was performed using a simple 8-colour panel to identify blasts at various timepoints and were compared to the initial unfixed flow data. We determined the recovery of the blast count and tracked changes in the intensity of leukemia-defining markers over the time course.

Results: Both Transfix and Streck fixatives were well suited to maintaining blast counts for at least three days. While Transfix was generally superior in maintaining the median fluorescence intensity (mfi) of leukemia markers, both preservatives functioned well to stabilize the cell counts. Streck, already used at St Jude for cell-free DNA applications, had the advantage of being able to stabilize cell counts at room temperature and to maintain that stability for at least six days. Transfix-treated CSF samples must be stored at 4oC and the stability is only maintained for three days.

Conclusions: We have identified a suitable fixative for CSF flow cytometry applications. We plan to incorporate Streck preservative into the clinical flow laboratory workflow for CSF analysis.

Presenter: Shaohua Lei, PhD

Mentor: Charles G. Mullighan, MBBS, MD

Co-Authors: Sujuan Jia, Sunitha Takalkar, Ti-Cheng Chang, Xiaotu Ma, Karol Szlachta, Ke Xu, Zhongshan Cheng, Yawei Hui, Selene C. Koo, Paul E. Mead, Qingsong Gao, Priyadarshini Kumar, Colin P. Bailey, Jobin Sunny, Alberto S. Pappo, Sara M. Federico, Giles W. Robinson , Amar Gajjar, Jeffrey E. Rubnitz, Sima Jeha, Ching-Hon Pui, Hiroto Inaba, Gang Wu, Jeffery M. Klco, Ruth G. Tatevossian, Charles G. Mullighan

Genomic profiling of circulating tumor DNA for childhood cancers

Abstract
The utility of circulating tumor DNA (ctDNA) analysis has not been well-established for disease detection and monitoring of childhood cancers, especially leukemias. We developed PeCan-Seq, a deep sequencing method targeting diverse somatic genomic variants in cell free samples in childhood cancer. Plasma samples were collected at diagnosis from 233 children with hematologic, solid and brain tumors. All children with hematologic malignancy (n=177) had detectable ctDNA at diagnosis. The median ctDNA fraction was 0.77, and 97% of 789 expected tumor variants were identified, including sequence mutations, copy number variations, and structural variations responsible for oncogenic fusions. In contrast, ctDNA was detected in 19 of 38 solid tumor patients and 1 of 18 brain tumor patients. Somatic variants from ctDNA were correlated with minimal residual disease levels as determined by flow cytometry in serial plasma samples from patients with B-cell acute lymphoblastic leukemia (B-ALL). We showcase multi-tumor detection by ctDNA analysis for a patient with concurrent B-ALL and neuroblastoma. In conclusion, PeCan-seq sensitively identified heterogeneous ctDNA alterations from 1mL plasma for childhood hematologic malignancies and a subset of solid tumors. PeCan-seq provides a robust, non-invasive approach to augment comprehensive genomic profiling at diagnosis and mutation- specific detection during disease monitoring.

Presenter: Lindsey Montefiori, PhD

Mentor: Charles G. Mullighan, MBBS, MD

Co-Authors: Huimei Wei, Josi Lott, Chun Shik Park, Yunchao Chang, Laura Janke, Charles Mullighan

A mouse model of ectopic BCL11B activation to investigate lineage ambiguous leukemia

Abstract
The T-lineage transcription factor gene BCL11B is aberrantly activated in a subset of lineage ambiguous leukemias with T and myeloid features. Despite showing a T-lineage immunophenotype, these leukemias resemble immature hematopoietic stem and progenitor cells at the transcriptional level and display immunophenotypic plasticity, suggesting an origin in a primitive or otherwise non-T hematopoietic cell type. How ectopic BCL11B functionally interacts with the cell of origin to drive lineage plasticity and transformation are unknown. To investigate these questions, we developed a mouse model of inducible BCL11B expression in phenotypically-defined cell types using the Cre-Lox system, including hematopoietic stem cells (HSCs, Vav-Cre and Scl-CreERt), committed B cells (CD19-Cre), and committed myeloid cells (LysM-Cre). Constitutive BCL11B activation in HSCs is embryonic lethal. Analysis of embryonic day 12.5 fetal liver cells showed an expansion of the HSC compartment, yet these cells failed to expand ex vivo, suggesting a functional defect that might explain the embryonic lethality. Conditional BCL11B activation in HSCs of 6 week-old mice was similarly lethal, although the same HSC expansion was not observed and the cause for this lethality remains under investigation. In CD19+ B cells, ectopic BCL11B blocked B cell development; in contrast, BCL11B was tolerated in committed myeloid cells, with preliminary data suggesting that ectopic BCL11B may promote a pathologic expansion of the CD11b+ compartment. Thus, our initial characterization of this mouse model demonstrates that ectopic BCL11B imparts a range of phenotypic consequences of varying severity depending on the developmental stage in which it is expressed. Ongoing work is interrogating the molecular basis of this context-dependent function using RNA-sequencing and chromatin binding assays, as well as testing BCL11B’s oncogenic capacity using transplant studies and the introduction of co-occurring mutations.

Presenter: Yunchao Chang, Ph D

Mentor: Charles G. Mullighan, MBBS, MD

Co-Authors: Yunchao Chang, Vanshita Goel, Gisele Nishiguchi, Sarah M. Young, Pankaj Ghate, Zhe Shi, Kevin McGowan, Anup Aggarwal, Jason Ochoada, Jeanine Price, Lei Yang, Francisca N. de Luna Vitorino, Renee Dean, Joanna K. Lempiäinen, Josi Lott, Suiping Zhou, Vishwajeeth Pagala, Yingxue Fu, Zuo-Fei Yuan, Anthony A High, Vibhor Mishra, Sandi Radko-Juettner, Baranda S. Hansen, Shondra M. Pruett-Miller, Benjamin A. Garcia, Marcus Fischer, Martine F Roussel, Zoran Rankovic, Charles G. Mullighan

Identification of a potent and selective PPIL4 degrader from a focused library of Cereblon modulators

Abstract
Targeted protein degradation (TPD), utilizing proteolysis targeting chimeras (PROTACs) and molecular glues (MGs), represents a groundbreaking strategy in drug discovery, offering solutions for targeting traditionally “undruggable” proteins. Previously, we reported the design, synthesis, and screening of a large, diverse library of Cereblon (CRBN) modulators against panels of acute leukemia (AL) and medulloblastoma (MB) cell lines. In this study, we identified SJ42872, a potent and selective degrader of peptidylprolyl isomerase-like 4 (PPIL4), demonstrating specificity over known IMiD neosubstrates such as GSPT1, CK1α, and IKZF1/3. SJ42872 displayed CRBN-dependent cytotoxicity in AL and MB cell lines, inducing cell cycle arrest and apoptosis. Knockdown of PPIL4 via gRNA further reduced cell growth in acute lymphoblastic leukemia cell lines MHH-CALL-4 and NALM-6 in cell fitness assays. Mutagenesis of four glycine residues in potential PPIL4 degrons identified a critical degron around G278, with overexpression of the PPIL4-G278N mutant conferring resistance to SJ42872 in both NALM-6 and MHH-CALL-4 cells. To better understand the interactions between PPIL4 and CRBN, we are pursuing the crystal structure of PPIL4 in complex with SJ42872 and CRBN. Additionally, ongoing pharmacokinetic/pharmacodynamic (PK/PD), tolerability, and efficacy studies aim to assess the therapeutic potential of SJ42872 in vivo. Collectively, these findings position SJ42872 as a valuable chemical probe for investigating the role of PPIL4 in both basic biology and drug discovery research.

Presenter: Yiming Wu, PhD

Mentor: Charles G. Mullighan, MBBS, MD

Co-Authors: Katarzyna Szoltysek, Tony Brown, Emily Backhaus, Amanda Brewer, Tanya Khan, Qingsong Gao, Keren Zhou, Yunchao Chang, Kathryn G Roberts, Kristine R. Crews, Ilaria Iacobucci, Ruth Wang'ondu, Jun Yang

Revolutionizing B-ALL Cell Line Development: Novel Generation to Uncover Therapeutic Vulnerabilities

Abstract
B-cell acute lymphoblastic leukemia (B-ALL) is the most common childhood cancer, with over 20 genomic subtypes. Cell line models are crucial for identifying vulnerabilities and developing precision therapies. However, most B-ALL cell lines are KMT2A-rearranged, BCR::ABL1+, TCF3::PBX1+ and MEF2D-r, leaving many subtypes unrepresented. Here, we developed a novel approach to generate B-ALL cell lines.

Previous methods attempted to adapt patient samples to grow in standard culture media, such as RPMI1640+FBS, but had low success rates. Here, we initiated culture from frozen patient-derived xenograft (PDX) cells. These were thawed and cultured with human mesenchymal stromal cells (MSCs) in serum-free medium, SFEM II, generally promoting expansion within a week. We then assessed serum and stromal cell dependency and further adapted PDX cells to stromal-free culture either with or without FBS. A new cell line was defined if it underwent exponential expansion for at least eight passages without co-culturing with MSCs. The cell lines were then cyropreserved and characterized with short tandem repeat analysis and whole-genome/transcriptome sequencing (WGS/WTS).

We first developed and tested the workflow with nine low hypodiploid (LH) and seven BCR::ABL1-like PDX samples, including CRLF2-r (N=4) and ABL1-r, EPOR-r and NTRK3-r (N=1 each) samples sourced from the St. Jude PDX repository PROPEL (propel.stjude.cloud). In the initial step, 10 samples proliferated on MSCs in SFEM II medium. In the serum and stromal cell dependency test, we found BCR::ABL1-like samples either grew without MSCs in medium with FBS or co- culture with MSCs in SFEM II medium, while LH samples only proliferated in SFEM II medium, with or without MSCs. Overall, two to eight months was required to generate a new cell line from PDX samples. In total, we generated eight LH and three BCR::ABL1-like cell lines out of 16 attempted, with their doubling time less than seven days.

We then tested the workflow with 17 PDX samples, including five NUTM1-r samples, four BCR::ABL1+ samples with either wild type or dominant negative (IK6) IKZF1 genotypes, two IKZF1 N159Y and three TCF3::HLF samples, and one sample each with near haploid, PAX5alt, PAX5 P80R, and ZEB2/CEBPE. NUTM1-r samples showed the lowest success rate, with only one cell line generated. Twelve cell lines were generated from the other 13 PDX samples, with one IKZF1 N159Y sample failing. This demonstrated our approach has a high success rate and is compatible with the majority of B-ALL subtypes. Notably, all these new cell lines consistently match the subtypes of PDX/original patient samples via t-SNE analysis, indicating they faithfully reflect the transcriptomic features of their corresponding subtypes. Additionally, we found BCR::ABL1+, BCR::ABL1-like, near haploid, and HLF-r samples grew in medium both with and without FBS, while other subtypes only grew in SFEM II medium. Notably, most samples required co-culture with MSCs or SFEM II medium, particularly subtypes not previously modeled, which in part explains the high success rate of this approach.

Finally, we studied the therapeutic vulnerabilities of these 24 new B-ALL cell lines by testing cytotoxicity of 36 commercial drugs. Drug sensitivity for specific subtypes was evaluated by comparing the average area under the dose-response curves (AUC) with those of other subtypes. LH cell lines (N=8) demonstrated resistance (average AUC ratio > 1.4) to inotuzumab, ponatinib, prednisolone, selinexor, and clofarabine, but showed sensitivity (average AUC ratio < 1.1 and average AUC for LH < 0.8) to buparlisib, carfilzomib, gilteritinib, and venetoclax. In BCR::ABL1+/BCR::ABL1-like samples with IK6 (N=3), cells exhibited sensitivity (average AUC ratio < 1.1 and average AUC for IK6 < 0.8) to apoptosis modulators (birinapant and venetoclax), proteasome inhibitors (bortezomib and carfilzomib), as well as buparlisib, daunorubicin, luminespib, and selinexor, compared to samples with wild-type IKZF1 (N=2). Additionally, we found that HLF samples (N=3) were highly sensitive (average AUC ratio < 0.8) to A-485, clofarabine, and trametinib.

In summary, we developed a new approach, generating 24 B-ALL cell lines spanning 11 subtypes. Notably, we established eight cell lines for the high-risk LH subtype, for which no models previously existed. These models provide an important resource for identifying dependencies and therapeutic vulnerabilities.

Presenter: Ruth Wang’ondu, MD, PhD

Mentor: Charles G. Mullighan, MBBS, MD

Co-Authors: Devanand Bondage, Qiong Zhang, Wojciech Rosikiewicz, Beisi Xu, Ti-Cheng Chang, Hongjiang Jin, Surbhi Sona, William Wu, Stanley Pounds, Yiming Wu, Shondra Miller, Qingsong Gao, Charles Mullighan

Unique super-enhancers associated with gene expression in IKZF1 N159Y leukemia

Abstract
Although the 5-year survival of pediatric patients with B-ALL exceeds 90%, relapse affects 10%-20% of these patients and is the leading cause of cancer-related death in children. Alterations in the tumor-suppressor gene IKZF1 and its product Ikaros, a transcription factor expressed in all hematopoietic cell lineages, are associated with gene expression changes, chromatin remodeling changes, and drug resistance.  he IKZF1 N159Y mutation is a B-ALL subtype-defining lesion associated with an intermediate risk of relapse lesion and with tandem duplication of the mutated exon 5: IKZF1 N159Y-TD. We hypothesized that the IKZF1 N195Y mutant drives leukemogenesis through patterns of chromatin occupancy and transcriptional deregulation distinct from other IKZF1 variants. To determine the role of chromatin remodeling in defining the unique gene expression pattern in IKZF1 N159Y leukemia, we identify genome-wide targets of wild-type and mutant Ikaros, EP300 (p300), H3K27Ac, H3K27Me3, H3K4Me3, and H3K36Me3 using chromatin immunoprecipitation and sequencing (ChIP-Seq) in an IKZF1 N159Y patient-derived xenograft cell line. IKZF1 occupancy was identified at 7604 chromatin regions (9.4% of all regions, 52.8% of regions associated with promoters) associated with promoters of 5850 genes. Of these promoters, 83% were co-occupied by H3K27Ac and 18.3% co-occupied  by H3K27Me3. Super enhancer (SEn) analyses identified significant changes in the number and identity of SEn compared to the DUX4 rearranged NALM6 cell line and to CD19 positive B Cells from healthy donors. The average number of enhancers (En) (2755) and SEn (214) were significantly decreased in the IKZF1 N159Y PDX cell line compared to NALM6 cell line (1313 SEn (P = 0.005), 21174 En (P = 0.003)) and CD19 positive B Cells from healthy donors (563 SEn (P=0.005), 5089 En (P<0.0001). More than 70% of reproducible SEn in the IKZF1 N159Y PDX cell line (137/186) are unique compared to the NAML6 cell line and 100% are unique compared to the SEn in CD19+ B cells from healthy donors. A significantly higher proportion of SEn (25.8%) were associated with upregulated (25.8%) vs downregulated (8.1%) differentially expressed genes in patients with IKZF1 N159Y subtype compared to CD19+ B cells from healthy donors (P<0.05). Highly ranked SEn included 12 SEn which are upregulated in patients with IKZF1 N159Y subtype relative to normal B-Cells and other leukemia. Six of the  12  SEn are associated with RHOB, SLC7A5, PRKCA, SALL1, GPER1,   and TGFB3 genes involved in lipid metabolism and/or TGFβ signaling. Our findings identify unique super-enhancers associated with gene expression in IKZF1 N159Y leukemia. Future studies will investigate the role of IKZF1 N159Y-induced super-enhancer reprogramming in leukemogenesis and B-cell differentiation.

Presenter: Bisi Miao, PhD

Mentor: Jian Xu, PhD

Co-Authors: Jaquelyn Zoine, Yuannyu Zhang, Xiaofei Gao, Wenhuo Hu, Naga Sarada Achyutuni, Massayuki Umeda, Jing Ma, Ilaria Iacobucci, Kim Nichols, Mitchell J. Weiss, Charles G. Mullighan, Jeffery Klco, Paulina Velasquez, Jian Xu

Enhancer Hijacking-Mediated Transcriptional Competition Drives Immune Gene Dysregulation During Leukemogenesis

Abstract
Emerging evidence underscores a critical role for enhancer hijacking, the repositioning of a transcriptional enhancer to the proximity of a proto-oncogene by genomic structural variants (SVs), in driving cancer development, particularly in leukemia and other cancer types. Despite these findings, two key questions remain difficult to address: First, how does enhancer hijacking deregulate transcription programs to promote oncogenesis? Second, since enhancer hijacking does not result in the formation of fusion oncoproteins— typically targeted in therapeutic approaches—a critical question is how to target oncogenic programs driven by enhancer hijacking?

To address these questions, we focus on the second most common SV in infant AML, t(7;12)(q36;p13), characterized by aberrant activation of the homeobox transcription factor MNX1. Using a sequence-based deep learning model, we identify ectopic enhancer-promoter loops between ETV6 and MNX1 loci in t(7;12) AML cells. CRISPR editing of t(7;12) in K562 cells results in allele-specific activation of MNX1 and silencing of ETV6. Through HiChIP and enhancer-targeting CRISPRi assays, we provide direct evidence that t(7;12) leads to the repositioning of ETV6 enhancers to activate MNX1 transcription. Importantly, we observe that MNX1 and ETV6 compete for chromatin binding at both the hijacked enhancers and genome-wide in CRISPR-edited t(7;12) cells. Using dTAG- mediated acute degradation in cord blood CD34+ hematopoietic stem/progenitor cells (HSPCs), we show that MNX1 and ETV6 act in a competitive and antagonistic manner to regulate chromatin binding and gene transcription, resulting in dysregulated immune gene expression. Furthermore, MNX1 enhancer hijacking- induced transcriptional competition leads to increased expression of CAR antigen genes, including CD33 and CD70, in t(7;12) AML cells, whereas CD33- or CD70-targeting CAR-T cells effectively eliminate t(7;12) AML cells in patient-derived xenografts. Together, our findings uncover new insights into the role of enhancer hijacking- induced transcriptional competition in leukemia pathophysiology, and propose a new therapeutic strategy for AML subtypes driven by transcriptional competition and immune gene dysregulation.

Presenter: Jin Wang, PhD

Mentor: Jian Xu, PhD

Co-Authors: Wenhuo Hu, Junhua Lyu, Aditya Sheth, Hui Cao, Hieu S Vu, Xiaofei Gao, Min Ni, Jian Xu

Abstract
Mesenchymal stromal cells play crucial roles in regulating hematopoietic stem cell quiescence and regeneration in the bone marrow microenvironment. They also interact with evolving leukemia cells and undergo molecular and functional alterations during the progression of hematologic malignancies. However, the dynamic interactions between leukemia cells and stromal cells, and their contributions to leukemia pathophysiology, remain largely unknown. Here, we established multiple inducible acute myeloid leukemia (AML) mouse models and performed single-cell multi-omic profiling to investigate the dynamic interactions between bone marrow stromal cells and both leukemic and non-leukemic hematopoietic cells across emergent, progressive, and terminal stages of leukemia progression. We identified a new population of stromal cells, characterized by enhanced energy metabolism, cell proliferation, and inflammatory signaling, in the bone marrow of leukemic mice. These cells are induced by evolving AML cells and expand rapidly, becoming the dominant fraction of bone marrow stroma during late-stage leukemogenesis. We observed rewired energy metabolism in this population, driven by altered fatty acid metabolism, which may contribute to impaired osteogenesis and increased inflammatory signaling. Analysis of cell-cell communication between stromal and hematopoietic cells revealed diminished support for normal hematopoietic cells but increased support for leukemic cells. Therefore, the time- resolved, single-cell profiling of the bone marrow microenvironment provides evidence that leukemic cells interact with and actively remodel the bone marrow stroma by reshaping its metabolic programs to promote leukemia progression. These findings provide new insights into the role of leukemia-microenvironment crosstalk in disease pathobiology and suggest new opportunities to intervene in the progression of hematologic malignancies.

Presenter: Syed Farhan Ahmad, PhD

Mentor: Jian Xu, PhD

Co-Authors: Wenhuo Hu, Kaili Wang, Yuannyu Zhang, Jian Xu

Elucidating the role of genomic repetitive element-mediated viral mimicry response in leukemia therapy

Abstract
Hypomethylation agents (HMA), including 5-azacytidine (AZA) and decitabine (DAC), are the first-line treatment for myelodysplastic syndromes and myeloid leukemias. HMA-induced DNA hypomethylation can lead to the reactivation of tumor suppressor genes, thereby impairing leukemia cell proliferation and survival. However, the broader transcriptional impact of HMA treatment, especially on epigenetically silenced non-coding repetitive elements, and its contribution to therapeutic outcome remain largely unexplored. We hypothesize that HMA-mediated reactivation of retrotransposable elements (RTEs) triggers an antiviral-like immune response, known as ‘viral mimicry’, in leukemia cells, contributing to the therapeutic efficacy of hypomethylating agents. To test this hypothesis, we performed transcriptomic and epigenomic profiling of eight pediatric and adult myeloid leukemia cell lines at multiple time points (day 1, 4, and 7) following AZA or DAC treatment and developed computational pipelines to annotate and quantify RTE expression. We identified a subset of RTEs, including endogenous retroviral elements (ERVs), that are reactivated in a cell type-, oncogenotype-, and/or treatment-specific manner. RTE reactivation was associated with dysregulated immune gene signatures, cellular stress, and apoptosis in leukemia cells. Furthermore, by analyzing genomic locus-specific RTE expression, we uncovered distinct features, such as epigenetic state, evolutionary age, and developmental regulation, that can explain HMA-mediated effects on RTE reactivation. Taken together, an in-depth analysis of the RTE-mediated viral mimicry response will not only elucidate the molecular mechanisms underlying HMA-mediated transcriptional reprogramming in leukemia cells but also establish RTE-induced viral mimicry response as a potential biomarker for hypomethylation therapy.

Presenter: Junhua Lyu, PhD

Mentor: Jian Xu, PhD

Co-Authors: Wenhuo Hu, Hieu S. Vu, Hui Cao, Gen Zhang, Feng Cai, McKenzie Patrick, Jin Wang, Kaili Wang, Yuannyu Zhang, Aditya Sheth, Stephen S. Chung, Ralph J. DeBerardinis, Min Ni, Jian Xu

Metabolic Compartmentalization Regulates Hematopoietic Stem Cell Function and Leukemogenesis

Abstract
Branched-chain amino acids (BCAAs), including leucine (Leu), isoleucine (Ile), and valine (Val), are essential amino acids required for normal development. The levels of BCAAs are regulated by branched-chain amino transferase (BCAT) isozymes, including the cytosolic BCAT1 and mitochondrial BCAT2, which catalyze the first step of BCAA catabolism. Despite their homology, BCAT1 and BCAT2 are segregated into distinct subcellular compartments and tissues. BCAT2, but not BCAT1, mediates BCAA metabolon formation in the mitochondria for BCAA oxidation and homeostasis, whereas the overexpression of cytosolic BCAT1, but not mitochondrial BCAT2, promotes oncogenesis in various cancer types, including acute myeloid leukemia (AML). Despite these findings, the role of BCAT isozyme-mediated compartmentalized BCAA metabolism in hematopoietic stem cell (HSC) function and leukemogenesis remains unexplored. To address this, we generated mouse models with hematopoietic-specific knockout or knock-in overexpression of BCAT1 or BCAT2. We found that BCAT1 is dispensable for normal hematopoiesis, but its overexpression selectively enhances HSC function. In contrast, BCAT2 is required for hematopoiesis, but its overexpression has no effect on HSCs. Importantly, the BCAT isozyme-mediated effects on HSC function are compartment-dependent, with BCAT1 and BCAT2 regulating compartment-specific metabolic programs that support HSC function. Furthermore, elevated BCAT1 expression correlates with poorer survival in AML patients with specific mutations (e.g., FLT3-ITD). Consistent with the oncogenic role for BCAT1 in myeloid leukemia, we found that BCAT1 overexpression accelerates disease progression in FLT3-ITD mutant knock-in mice and PDX models. Mechanistically, BCAT1 promotes leukemogenesis by activating MAT2A-mediated methionine metabolism and SAM-dependent epigenetic processes. Inhibition of MAT2A with AG-270, a first-in-class, orally bioavailable MAT2A inhibitor, selectively targets leukemia-initiating cells in preclinical models. Thus, the compartmentalized metabolism of BCAAs regulates HSC function and leukemogenesis, highlighting selective metabolic vulnerabilities of leukemia- initiating cells.

Presenter: Wenhuo Hu, PhD

Mentor: Jian Xu, PhD

Co-Authors: Syed Farhan Ahmad, Jian Xu

Hematopoietic Development-Related Transposon Elements Associated with Leukemia Patient Survival

Abstract
Transposable elements (TEs), or “jumping genes”, exhibit developmental stage- and lineage-specific expression patterns. However, due to the inherent technical challenges in studying repetitive elements, the expression landscape of TEs in normal hematopoietic stem and progenitor cells (HSPC) and in leukemia cells, along with their role in disease pathogenesis and prognosis, remain largely uncharacterized. In this study, we developed TE-customized computational pipelines to integrate transcriptomic and epigenomic data, and discovered distinct TE expression patterns in HSPCs, including TE subfamilies and genomic loci uniquely expressed in hematopoietic stem cells (HSC), each lineage-restricted progenitors (e.g., CMP, GMP, and MEP), and mature lineages (e.g., erythroblasts and megakaryocytes). Most of the developmental stage- or lineage-specific TEs are located close to or within the annotated protein-coding genes. By integrating RNA-seq and ATAC-seq datasets, we found that a subset of TE host genes displayed significant changes in gene expression and chromosome accessibility during HSC differentiation. Among them, IFNGR1, which plays a critical role in immune regulation and HSC maintenance, was highly expressed in HSCs compared to differentiated progenitors. The expression of IFNGR1 is accompanied by the expression and chromatin accessibility of an intronic TE element, suggesting the role of TE in regulating host gene expression critical for HSC function. We further quantified TE expression in multiple AML cohorts, including TCGA, Beat AML, and TARGET. We found that chromosomal cytogenetic changes, such as MLL rearrangements, AML-ETO, CBFB-MYH11, and PML-RARA, were associated with increased TE expression. In contrast, other recurrent oncogenic mutations, such as DNMT3A, IDH1/2, and TET2, were associated with lower TE activation. We further identified a subset of myeloid progenitor-associated TE subfamilies, including LINE1 retrotransposons, to be associated with patient outcomes, indicating the potential of developmentally regulated TEs as prognostic biomarkers in AML. Taken together, our findings underscore the significance of comprehensive annotation and quantification of TE expression in normal and malignant hematopoiesis, with translational implications for identifying TE subfamilies as potential biomarkers in AML and other leukemia subtypes.

Presenter: Xiaofei Gao, PhD

Mentor: Jian Xu, PhD

Updates on the Advanced Technology and Genomics (ATG) Core

Abstract
The Advanced Technology and Genomics (ATG) core within the Center of Excellence for Leukemia Studies (CELS) aims to leverage the technological advances, the collective expertise of our team, and the collaborative environment at St. Jude to facilitate the development of innovative technologies that enhance research, diagnosis, prognosis, and treatment of childhood cancers. To achieve this goal, we are developing a streamlined, interactive, and productive platform for innovation in genomic and metabolomic technologies, proof-of-principle validation, and project-based collaborations. In this update, we provide an overview of the major accomplishments and ongoing development of core technologies and collaborative projects. We have established effective collaborative workflows for genomics and metabolomics studies across the institution. The collaboration involving LAMPS (Light Activated Macromolecular Phase Separation) technology to analyze the molecular composition of fusion oncoprotein (FO)-mediated chromatin regulation has contributed to an NIH P01 grant application. The ongoing MAP (Metabolomic Analysis of Patient-Derived Xenografts) project has provided valuable insights into understanding the metabolic alterations in pediatric blood cancers. Additionally, we developed a metabolomics workflow that now supports 48 projects across 12 labs at St. Jude. Since September 2023, it has processed 3,962 samples with equipment usage exceeding 90%. These efforts have led to two publications (Fan et al., 2024, Hepatology; Lyu et al., 2024, Science) and five additional manuscripts, highlighting the platform's productivity and impact. Through technology development and collaborative efforts, we envision the ATG core as an “incubator” for innovative, user-customized methodologies, a catalyst for applying cutting-edge technologies to childhood cancer research, and a platform that synergizes team science.