TET 2 Alterations in Myeloid Malignancies, Impact on Clinical Characteristics, Outcome, and Disease Predisposition

TET2 mutations are the most common somatic genetic lesions in myeloid neoplasms. TET2 mutant clones have been found also in healthy individuals, increase with age, and convey an increased risk for myeloid clonal diseases. The TET2 gene is very polymorphic, with hundreds of single nucleotide polymorphisms (SNPs) of unknown clinical impact, but with some variants that may be pathogenically important. Similarly, somatic mutations affect all portions of the gene, and can be missense or truncating, homo-, hemi- and heterozygous. While the majority of TET2 mutations are ancestral, they can also be subclonal, implicating the clonal architecture in the consequences of TET2 lesions. This diversity may be hampering establishment of the clear prognostic impact of TET2 mutations.

Taking advantage of a large cohort of patients (pts, N=4985 including 1616 MDS, 871 MDS/MPN, 1782 pAML, 304 sAML, 333 MPN, and 79 therapy-related MDS/AML/MDS-MPN) analyzed by targeted deep sequencing for TET2 and other common myeloid lesions, we examined the distribution and impact of TET2 mutations. DNA sequencing of all coding exons of TET2 and 61 other genes representing the most common somatic mutations in myeloid neoplasms. Nonsynonymous alterations not present in SNP database (dbSNP) were annotated as somatic mutations or SNPs if present in myeloid and T cells whenever available. Nonsynonymous alterations not in dbSNP or ExAC databases and not confirmed to be somatic were excluded. Overall, TET2 somatic mutations (TET2^mut) were present in 920 pts (18%); 38% of MDS/MPN, 19% pAML, 16% MPN, 16% sAML, 12% MDS, and 13% of therapy related MDS/AML/MDS-MPN. Mutations included 16% missense, 33% frameshift deletions, 18% frameshift insertions, and 33% nonsense. TET2^mut pts were older than those with TET2 wild type (TET2^wt, 72 vs. 67 yrs, p<.001), had a higher presenting WBC (6 vs. 4 x10³ /uL, p <.001), and lower blast % (3 vs. 7%, p =.03). Similar findings were observed in each myeloid subtype. Overall, median OS for TET2^mut pts was similar to TET2^wt (12 vs. 17 mo, p =.20). Median OS was similar in TET2^mut pts compared to TET2^wt in pts with MDS (23 vs. 23 mo, p =.77), MDS/MPN (15 vs. 21 mo, p=.1), pAML (9 vs. 14 mo, p =.77), sAML (6 vs. 9 mo, p =.07), and MPN (30 vs. 35 mo, p =.66). Neither the type of mutation (mis-, nonsense vs. truncating) nor location (catalytic domain vs. other) impacted the OS. Using variant allelic frequencies (VAF), we established a clonal hierarchy in individual cases; 24% of TET2 mutations were ancestral, 17% subclonal, and 59% codominant. TET2 mutations were ancestral in 23% of MDS samples, 29% of MDS/MPN, 25% of pAML, and 19% of sAML. Whether the mutation was ancestral or subclonal did not impact OS. The presence of TET2^mut was associated with different mutations in each myeloid subtype. In MDS, TET2^mut were associated with APC (p<.001), ASXL1 (p<.001), BCOR (p<.001), BCORL1 (p<.001), ETV6 (p<.04), SUZ12 (p<.001), RAD21 (p<.02), NF1 (p<.001), KDM6A (p<.001), ZRSR2 (p<.001), and U2AF1 (p=.02), in MDS/MPD correlated with ASXL1 (p<.04), NRAS (p<.02), and SRSF2 (p<.05), in pAML with JAK2 (p<.001), RUNX1 (p =.05), and CBL (p=.05), and in sAML with RUNX1 (p<.001), ASXL1 (p<.001), BCORL1 (p=.01), SUZ12 (p=.02), STAG2 (p=.05), and JAK2 (p<.001).

When we next focused on germ line variants, we identified 2518 SNPs of TET2. All recurring SNPs were ranked according to the difference in their frequencies between pts and healthy controls. A large number of these SNPs were more common in our pts compared to controls, among them we identified 2 SNPs (both located in the dioxygenase domain) with a significantly higher frequency: SNP1 (OR 10.6, p<.0001), and SNP2 (OR 6.7, p=.02). We further investigated whether these SNPs were mutually exclusive or increased the risk for acquisition of somatic TET2 mutations; 91% of cases with SNP1 and 67% with SNP2 also acquired somatic mutations in TET2. In silico and crystallographic analyses showed that SNP1 is adjacent to the iron binding site (7^th beta stand in the jelly roll motif) and is predicted to change the orientation of a-KG binding and thereby to be hypomorphic. SNP2 is located in a hot spot area known to be targeted by 3 recurrent somatic mutations.

In conclusion, both somatic mutations as well as germ line variants affect TET2 in myeloid neoplasia. The interaction between clonal mutations and germ line lesions may lead to gain of function and thus a growth advantage. These mechanisms are currently being explored.