Comprehensive Genomic Profiling of Pediatric Therapy-Related Myeloid Neoplasms Identifies Mecom Dysregulation to be Associated with Poor Outcome

Therapy-related myeloid neoplasms (tMN) occur in children secondary to cytotoxic therapies used to treat pediatric malignancies, are typically resistant to conventional chemotherapy, require hematopoietic cell transplantation as the only curative option, and have a dismal prognosis. The genomic alterations that drive tMN in children have yet to be comprehensively described, and it is unclear if particular genomic lesions hold prognostic value. We have characterized the genomic profile of 62 pediatric tMN cases (tMDS: n=23, tAML: n=39) obtained from the St. Jude Children's Research Hospital Tissue Bank from patients diagnosed between 1987 and 2018. These cases arose following treatment for a variety of primary tumors (hematological (74%), bone and soft tissue (23%), and brain (3%)). Acute lymphoblastic leukemia was the most frequent primary tumor (n=39, 63%). Conventional cytogenetics (n=60) showed a complex karyotype (≥3 structural alterations) in 19 (32%) cases, and 7 of these cases contained a deletion involving chromosome 7 (del(7)). Eleven (18%) other cases without complex karyotypes had del(7). Deletions of chromosome 5 were present in 9 (15%) cases, but only in the context of a complex karyotype. We hypothesized that the patients' younger age and the different spectrum of primary tumor types and chemotherapy would give rise to a mutational spectrum distinct from adult tMN.

We used whole exome (WES), whole genome (WGS), and RNA sequencing (RNA Seq) to describe the mutational profile of our pediatric tMN cohort. WES was completed for 58 tumor/normal pairs using Nextera Rapid Capture Expanded Exome (Illumina). Fifteen cases were analyzed by WGS (11 also had WES). Normal comparator genomic DNA was obtained from flow-sorted lymphocytes. An average of 21 coding variants/patient (range: 1-131) was observed from the gene-coding region, and these include synonymous, non-synonymous, and splice site variants. Ras/MAPK pathway mutations were present in 44% of the cases (43 mutations in 27 cases). Canonical KRAS (n = 16), NF1 (n = 10), and NRAS (n = 7) mutations were the most frequent coding mutations. Eleven (18%) patients had either heterozygous deletion or a copy neutral loss of heterozygosity event involving chromosome 17p and the TP53 locus; 5 of these cases had concurrent TP53 missense mutations identified at allele frequencies near 100%. Unlike tMN in adults, mutations in PPM1D were not identified. RNA-Seq completed on 56 evaluable cases identified 28 (50%) cases with KMT2A rearrangement (KMT2Ar). MLLT3 was the most common fusion partner (n=13, 46%).

In addition to KMT2A rearrangements, RNA-Seq also identified a RUNX1-MECOM fusion. Alterations involving the MECOM locus have been described in some myeloid neoplasms like tMN, and its overexpression is associated with a poor prognosis and some AMLs with KMT2Ar. MECOM expression levels were variable in this cohort (FPKM range: 0.004 - 38.4) with 24 cases (43%) having an FPKM>5 (MECOM^High). In addition to the RUNX1-MECOM event, these 24 MECOM^High cases included 18 with KMT2Ar (64% of KMT2Ar group) and 1 with a NUP98 fusion (NUP98-HHEX). The remaining 4 MECOM^High cases demonstrate allele-specific MECOM expression, suggesting a cis-regulatory element is driving this expression. Two of these 4 cases have WGS and were found to contain a t(2;3)(p21;q26.2) involving MECOM on chromosome 3 and noncoding regions of chromosome 2 adjacent to ZFP36L2, a gene highly expressed in hematopoietic cells. ENCODE data supports that this region of the genome is an active enhancer in hematopoietic cells, suggesting a proximity effect in which this enhancer has been hijacked to drive high levels of MECOM expression. In our cohort, MECOM expression levels are predictive of a worse outcome (overall survival (OS) at 2 years: High=14.6% vs. Low=46.3%; log rank p<0.01). Although KMT2Ar was frequently present in our cohort and enriched in the MECOM^High group (High=75% (18/24) vs. Low=31% (10/32); p<0.01), high MECOM expression did not confer a significant survival difference within the KMT2Ar group (OS at 2 years: High=16.7% vs Low=40%; log rank p=0.33). Further, the presence of a KMT2Ar or a complex karyotype did not significantly affect the OS in this cohort.

In conclusion, we report the genomic profile of a large cohort of pediatric tMN cases and show that high levels of MECOM expression, a portion of which is driven by enhancer hijacking, predicts a worse outcome.