NF1 Tumor Suppressor Gene Inactivation in Juvenile Myelomonocytic Leukemia: New Genetic Evidence and Consequences for Diagnostic Work-up

Neurofibromatosis type 1 (NF-1) predisposes to juvenile myelomonocytic leukemia (JMML) via loss of function of the NF1 tumor suppressor gene and consecutive deregulation of Ras signal transduction. Affected individuals usually carry one defective NF1 allele in the germline; somatic inactivation of the second NF1 allele in hematopoietic cells is associated with transformation to leukemia. We previously demonstrated that a major mechanism for biallelic loss of NF1 function in patients with JMML/NF-1 is mitotic recombination leading to uniparental disomy (UPD) of the 17q chromosome arm (Flotho, 2007; Steinemann, 2010). Using contemporary resequencing and microarray technology, we have now revisited the genetics of NF1 inactivation in JMML. Specifically, we addressed two questions: 1) Are genetic findings in leukemic cells of JMML/NF-1 patients consistent with the clinical diagnosis and the two-hit concept? 2) Does the quintuple-negative (QN) group of JMML (patients without clinical evidence of NF-1 and negative for mutations in PTPN11, KRAS, NRAS, or CBL) contain unrecognized cases likely driven by NF1?

We investigated 156 children with JMML registered in studies EWOG-MDS 98 or 2006 and tested for mutations in PTPN11, KRAS, NRAS, and CBL. Twenty-five children (16%) were clinically diagnosed as NF-1 based on >=6 café-au-lait spots (CALS) or family history plus CALS. Family history was positive in 9 JMML/NF-1 patients; >=6, 4, and 1 CALS were described in 23, 1, and 1 patients, respectively. The median age at diagnosis of JMML in the NF-1 group was 35.9 months (range, 4.2 to 71.4). Sixteen children (10%) were grouped as JMML-QN. Granulocyte DNA from bone marrow or peripheral blood collected at time of diagnosis was used for next-generation sequencing of the entire NF1 coding sequence (custom Ampliseq enrichment and Miseq, Illumina). Pathogenicity of NF1 variants was assessed according to ACMG criteria. Affymetrix Cytoscan HD array analysis was applied to detect segmental deletions or copy number-neutral loss of heterozygosity (LOH).

Among 25 JMML/NF-1 cases, 8 exhibited an NF1 loss-of-function mutation at near-100% variant allelic frequency (VAF) in combination with UPD involving almost the entire 17q arm, suggesting single mitotic recombination as the leukemic driver. One case had an NF1 mutation at near-100% VAF and segmental 17q UPD, suggesting the unusual occurrence of double mitotic recombination. Nine cases carried two independent pathogenic NF1 mutations at near-50% VAF each; here, germline and somatic events could not be distinguished due to unavailability of non-hematopoietic or parental DNA. Four cases exhibited an NF1 microdeletion in combination with a pathogenic NF1 mutation at near-100% VAF; non-hematopoietic tissue available in 2 of these 4 cases failed to display the mutation, indicating the microdeletion as the constitutional event. A deleterious mutation at 71% VAF but no LOH was revealed in one sample. Only monoallelic evidence of NF1 deficiency was found in 2 cases.

In the JMML-QN group, 9/16 cases had previously unrecognized NF1 alterations. Compound-heterozygous pathogenic NF1 mutations were found in 2 and homozygous pathogenic NF1 mutations combined with focal LOH in 3 patients, strongly suggesting NF1 as the JMML driver. In the absence of clinical NF-1 features, the findings in these 5 children may be explained by postzygotic NF1 mosaicism, onset of JMML before clinical manifestation of NF-1, or double somatic NF1 hits in the hematopoietic lineage. Four cases carried monoallelic pathogenic or likely pathogenic NF1 mutations with VAF at ~50% or less, providing inconclusive evidence of driver function as it is also possible that these are secondary hits in a myeloproliferative neoplasm (MPN) driven by an unidentified event. There was no genetic evidence of NF1 involvement in the remaining 7 JMML-QN cases.

Several important conclusions can be drawn from our study: The clinical diagnosis is reliable in children with JMML/NF-1; we propose that it can be made on the basis of CALS and JMML alone as the patients are generally too young to display the conventional spectrum of NF-1 features. Children without features of NF-1 should only be assigned to JMML-QN after genetic work-up of NF1 because this will unmask involvement of NF1 in a significant number of cases. In suspected JMML-QN without identifiable NF1 lesion, other forms of MPN should be considered.