Clinical and molecular characterization of pure erythroid leukemias and sarcomas driven by nfia::CBFA2T3

Abstract Pure erythroid leukemias (PEL) and sarcomas driven by NFIA::CBFA2T3 are rare and aggressive malignancies. They most often present in the central nervous system (CNS) where they often masquerade as other neoplasms and are often refractory to standard chemotherapy regimens, altogether underscoring the need to establish an accurate molecular diagnosis and to identify novel therapeutic approaches. Here, we describe five pediatric cases of NFIA::CBFA2T3 or NFIA::CBFA2T2 PEL that each underwent complete clinical and molecular characterization including cytogenetics, targeted DNA sequencing, and bulk whole transcriptome sequencing (RNA-seq). All cases occurred in young children ages 5 months to 3 years of age at diagnosis. Four of five cases presented with either CNS or periorbital involvement, with the periorbital case also presenting with an intra-abdominal tumor and bone marrow involvement. The fifth case, surprisingly, only involved the bone marrow and not the CNS. Each patient received various chemotherapy regimens with variable responses. One case was refractory to multiple courses of both standard chemotherapy and targeted therapies. For two of the cases, complete remission was achieved and the patients underwent allogeneic hematopoietic stem cell transplantation (HSCT), but both patients eventually relapsed and died of their disease. The two remaining cases only have short-term follow-up to date. Targeted sequencing did not reveal TP53 mutations, which are often described in adult cases of PEL. Instead, we found frequent co-occurring mutations in signaling and epigenetic genes, analogous to most cases of acute myeloid leukemias (AML). Signaling genes with pathogenic mutations included NRAS, KRAS, BRAF, JAK1, JAK2, EPOR, and CRLF2. Epigenetic genes with pathogenic mutations included TET2, SETBP1, KMT2D, and ARID1A. Given that our five cases of NFIA::CBFA2T3/2 PEL had similar genetic features to AML, we performed and/or analyzed RNA-seq for all five cases and compared them to 1,503 cases of pediatric AML that also underwent RNA-seq. Intriguingly, we found that the transcriptional profiles for all five cases clustered closely between cases of NUP98::KDM5A AML and CBFA2T3:GLIS2 AML based on UMAP analysis. As expected, differences from AML included upregulation of erythroid transcriptional programs and HOXB gene family members (rather than HOXA gene family members that are upregulated in NUP98 and KMT2A-rearranged AMLs). Finally, we incorporated RNA-seq data from normal bone marrow (NBM) and CD34 enriched peripheral blood stem cells (PBSC) and performed target discovery analysis. We found that traditional immunotherapy targets (CD33 and CD123) were downregulated in NFIA::CBFA2T3/2 PEL compared to AML, but that CD70 was uniquely upregulated in all five cases of NFIA::CBFA2T3/2 PEL. We have established and have serially passaged a patient derived xenograft (PDX) model of NFIA::CBFA2T3 PEL from a bone marrow specimen obtained at the time of relapse from the patient who presented with CNS and intra-abdominal tumor involvement. Of four engrafted secondary recipient mice, three developed ataxia and circling behavior 7-17 weeks after tail vein injection of PEL cells harvested from the bone marrow of a primary recipient. CNS involvement was confirmed by MRI. At necropsy, brains were covered in a thick substance containing PEL cells and abdominal tumors were found, recapitulating the clinical presentation of the original patient. The fourth mouse had low level bone marrow involvement at necropsy. RNA-seq comparing cells from our PDX model and source patient material revealed a close correlation of transcriptional profiles. We are performing preclinical therapeutic experiments to determine whether CD70 is a candidate target for NFIA::CBFA2T3 PEL. We believe that similar approaches to target discovery are needed, where existing therapeutic approaches are repurposed for this and other rare and aggressive malignancies.