The Genomic Landscape of Wilms' Tumor 1 (WT1) Mutant Acute Myeloid Leukemia

Mutations in tumor suppressor genes and oncogenes are both potentially therapeutically actionable in acute myeloid leukemia (AML). The Wilms' Tumor 1 (WT1) gene is located on 11p13 and encodes a zinc finger transcription factor which has been found to be overexpressed and mutated in AML. In normal development, WT1 is only expressed in a small subset of hematopoietic stem cells. While its overexpression suggests an oncogenic role, the invariable presence of mutations in the cysteine-histidine zinc finger domains indicates a tumor suppressor function, similar to that in WAGR syndrome/11p deletion syndrome in which it was first discovered. Like its unknown function in AML, the clinical significance and genetic associations of WT1 mutations have been also controversial. Although studies of WT1 mutations in AML have been conducted, the lack of solid clinical and molecular characterization of large WT1-mutant (WT1^MT) AML cohort has hampered its definition.

In this study, we took advantage of a compendia of genomic results from Cleveland Clinic and publicly available data of 2188 AML patients (primary (p)AML, n= 1636; secondary (s)AML, n= 433; therapy-related (t)AML, n= 119, excluding cases with acute promyelocytic leukemia, MLL-rearrangement, and core-binding factor AML). While several reports only focused on cytogenetic normal AML (CN-AML), which represented 61% of our cohort, we additionally included all other cytogenetic risk groups. In total, WT1 mutations were detected in 5% (114/2188) of patients. WT1 mutations were enriched in pAML (85%) compared to sAML (11%) and tAML (4%). Thirty-nine patients (13%) carried more than 1 WT1 mutation. WT1^MT were younger [59 vs 64 years, P=0.0002] and more often females (55% vs 45%, P=0.03) as compared to WT1 wild type (WT1^WT) patients. Univariate analyses of baseline parameters showed that WT1^MT AML had a more proliferative phenotype with a higher WBC [15.1 vs 9.5 x10⁹/L, P=0.03] and bone marrow blast percentages [73 vs 59%, P=0.002] and with lower platelet counts [44 vs 56 x10⁹/L, P=0.008] compared to WT1^WT cases. In the WT1^MT cohort, 70% had a normal karyotype, with complex karyotype being significantly less frequent vs WT1^WT patients [4 vs 16%, P=0.001]. The most common cytogenetic abnormalities in WT1^MT patients included +8 (8%) followed by -9/del(9q) (3%) and -7/del(7q) (3%). Only 1 patient carried inv(3)/t(3;3) or -17/del(17p). In sum, no statistical differences in cytogenetics were found between WT1^MT vs WT1^WT AML patients. Next, identified mutational signatures of WT1^MT patients. A panel of 44 myeloid genes and their hotspot configurations were selected according to their relevance in AML. In comparison to WT1^WT AML patients, multivariate analyses showed that WT1^MT patients had higher odds of biallelic CEBPA (12 vs 3%; P=0.009) and FLT3 internal tandem duplication mutations (FLT3^ITD, 31 vs 16%; P=0.01) but lower odds of SRSF2 mutations (2 vs 9%, P=0.04). Since FLT3^ITD has been previously described to be associated with WT1 mutations, we also focused on investigating whether mutations in the tyrosine kinase domain (TKD) were frequent in WT1^MT as well. Although we found increased percentages of FLT3^TKD (11%) among the WT1^MT patients compared to WT1^WT cohort (8%), this difference did not reach statistical significance. To uncover multifactor lesions (cytogenetic and/ or additional molecular lesions) of prognostic importance, we performed survival analyses. Although the combination of WT1 mutations and FLT3^TKD shortened overall survival (OS) by 2-times in WT1^MT patients vs WT1^WT cases with FLT3^TKD (23.7 vs 45.9 months), this result was not significant (P=0.1). In addition, the concurrent presence of other cytogenetic and molecular features didn’t reveal significant impact on OS.

In sum, using an adequately powered cohort, our study of the genomic landscape of WT1^MT AML patients identified its genomic associations and their clinical and prognostic inferences. The application of advanced machine learning methods to large datasets of WT1^MT AML patients might be crucial to capture the complex genomic interactions of WT1 gene in AML.