TET2 and MSH6: A Novel Role for MMR in the Pathogenesis of Myeloid Malignancies

TET2 mutations (TET2^MT) occur in up to 40% of myeloid neoplasms and increase linearly with age. That these lesions are founding hits can be concluded from clonal architecture and from the presence of subclinical TET2^MT clones in healthy individuals at risk for subsequent MDS. Though TET2^MT are heterogeneous, most disrupt TET2 catalytic domain dioxygenase activity; these widespread LOF mutations implicate TET2 as a bona fide tumor-suppressor gene (TSG). As downstream targets have not been identified as being associated with TET2^MT, the precise mechanisms of TET2^MT in leukemogenesis remains unclear. We studied whether TET2^MT predispose cells to acquiring additional oncogenic mutations through faulty DNA repair to produce clonally acquired mutator phenotypes.

In WES analyses of 435 myeloid neoplasia patient (pt) samples, we identified 95 with somatic TET2^MT enriched vs controls in the ExAC database, and used 6 bioinformatics tools to predict the effect on protein function based on algorithms incorporating sequence homology, amino acid properties, sequence context, and similarity to Mendelian diseases. Most pts (63) were predicted to have damaging TET2^MT in 3+ out of six of these tools.

To validate TET2^MT clinical relevance, we established an in-vitro 2D-UPLC-MS/MS assay to assess levels of the TET2-dependent DNA oxidation products 5hmC, 5fC, and 5CaC. While TET2 +/- knockout mouse samples had nonsignificant differences in 5hmC (1.2-fold lower levels vs TET2^WT, p=.23) and 5fC (1.4-fold increase vs TET2^WT, p=.07), they had significantly lower 5CaC (2.4-fold decrease vs TET2^WT, p<.0001). TET2 -/- mouse samples had significantly lower 5hmC (1.6-fold reduction vs TET2^WT, p<.0001), 5fC (1.3-fold reduction vs TET2^WT, p=.011), and 5CaC (2.4-fold reduction vs TET2 WT, p<.0001). Human MOLM-13 TET2 -knockdown (TET2 kd) cells also had 1.6-fold lower levels of 5hmC (p<.0001). Reassuringly, eight samples from TET2^MT patients also had lower 5hmC (5/8 samples, peak 2.9-fold reduction vs TET2^WT, p<.0001), 5fC (6/8 samples, peak 8.2-fold reduction vs TET2^WT, p<.0001), and 5CaC (5/8 samples, peak 2.6-fold reduction vs TET2 WT, p<.0001), and all 8/8 TET2^MT pt samples had significant declines in at least one of the three TET2-dependent oxidation vs TET2^WT samples.

To test whether TET2^MT predisposed cells to acquiring additional genomic mutations, we measured the total number of single nucleotide variants (SNV) in the entire exome in each pt sample, excluded pts with mutations in known DNA repair proteins, and compared the median WES SNV of TET2^MT pts to TET2^WT pts. Consistent with our theory, TET2^MT pts had a 1.5-fold increase in median WES SNV (p<.0001), particularly at 5hmC sites. Pts with a TET 2^MT variant allele frequency>50% had a 2.1-fold increase in median WES SNV (p=.03), consistent with a gene-dose effect.

TET2 -/- knockout mice had 1.4-fold increased mutagenicity at TET2-dependent active demethylation sites, in a manner suggestive of defective mismatch repair (MMR). HeLa TET2 kd cells also had a 24-fold increase in spontaneous mutations, reversed with TET2 cDNA knock-in. MMR gene expression (MSH2, MSH3, MSH6, and MLH1) was equivalent among 20 TET2^MT MDS pts, 71 TET2^WT MDS pts, and 17 controls, indicating MMR downregulation was not responsible for the hypermutagenicity. Five pts with TET2^MT were microsatellite-stable (MSS) at 5 TET2-independent poly-dA microsatellite loci, suggesting that hypermutagenicity was not driven by global MMR dysfunction and may occur only at CpG-containing microsatellites.

By overexpressing Flag/V5-tagged TET2 in mouse-MEL cells and subjecting them to protein affinity purification, we identified MSH6 as a novel TET2 binding partner. TET2:MSH6 interactions were validated in co-immunofluorescence experiments in MEL and MOLM-13 cells, and TET2kd altered MSH6 nuclear co-localization. MOLM-13 TET2kd increased PARP inhibitor sensitivity by almost 3-fold, suggesting that TET2kd renders cells vulnerable to perturbations in DNA repair pathways.

In sum, we uncovered novel connections among TET2, MSH6, epigenetic modifications, and genomic instability. Given that MSH6 is known to preferentially bind 5hmC, TET2 may target MSH6 to TET2-dependent DNA loci. Genomic instability due to TET2 dysfunction may allow therapeutic targeting of DNA repair proteins in the subset of hematologic malignancy pts with TET2^MT.