AUTHORS
Asmita Shukla, MD, Vikram Dhillon, DO, MBA2, Hunter Dlugas, MS3, Jeff Justin Aguilar, MD, MBA4, Dakshin Padmanabhan, MD5, Sushmitha Nanja Reddy, MD, MBBS6, Sikander Chohan, MD7, Markie Zimmer, MD8, Julie Boerner, PhD9, Jay Yang, MD10, David Carr11, Yang Shi, PhD12, Suresh Kumar Balasubramanian, MD13
AFFILIATIONS
1 Karmanos Cancer Institute, Detroit, Michigan, United States
2 Karmanos Cancer Institute, Detroit, Michigan, United States
3 Karmanos Cancer Institute, Detroit, Michigan, United States
4 DMC Sinai Grace, Detroit, Michigan, United States
5 Karmanos Cancer Institute, Detroit, Michigan, United States
6 Karmanos Cancer Institute, Detroit, Michigan, United States
7 Wayne State University, Detroit, Michigan, United States
8 Mayo Clinic, Rochester, Minnesota, United States
9 Karmanos Cancer Institute, Detroit, Michigan, United States
10 Karmanos Cancer Institute, Detroit, Michigan, United States
11 Karmanos Cancer Institute, Detroit, Michigan, United States
12 Karmanos Cancer Institute, Detroit, Michigan, United States
13 Karmanos Cancer Institute, Detroit, Michigan, United States
ABSTRACT
Introduction
DNMT3A is mutated in approximately 20% of myeloid neoplasms (MN) and carry a variable impact on clinical outcomes based on the mutational locus. Protein domain localization and mutational burden have each been independently associated with distinct disease features, but whether they jointly define biologically meaningful subgroups and transcriptional programs is unknown. We integrated clinical, genomic, and transcriptomic datasets to determine whether combining these features better subclassifies DNMT3A mutated MN.
Methods
We analyzed 2,998 patients with DNMT3A mutated MN from AACR GENIE, cBioPortal, dbGaP, and Karmanos Cancer Institute. Mutations were categorized by protein-domain localization (NTD, PWWP, ADD, MTase) and mutational complexity (R-882 as single canonical missense (CMS), non-R882 as single noncanonical missense (NCMS), double CMS, double NCMS, and mixed CMS/NCMS). Clinical characteristics, cytogenetics, co-mutation patterns, and OS were compared using chi-square and Kaplan-Meier analyses. RNA sequencing from an independent dbGaP cohort characterized transcriptional programs by subgroup; DE via DESeq2 (ashr shrinkage; |log2FC|>1, nominal P<0.05), GO/KEGG enrichment via hypergeometric test (BH-adjusted P<0.05, q<0.2).
Results
Among 2,998 DNMT3A mutated patients: 1,195 (40%) were single CMS, 1,488 (50%) single NCMS (NTD, n=85; PWWP, n=245; ADD, n=346; MTase, n=812), 39 (1%) double CMS, 236 (8%) double NCMS, and 40 (1%) mixed CMS/NCMS. Single CMS mutations were enriched in primary AML (pAML) 1028/1195 (86%) and frequently co-occurred with NPM1 681/1195 (57%) and FLT3-ITD 275/1195 (23%). Compared with single CMS, NTD domain of single NCMS was less frequent in pAML (37/85, 43% vs. 1028/1195, 86%; p<0.0001) but were more frequent in MDS (17/85, 20% vs. 35/1195, 3%; p<0.0001) and MPN (5/85, 6% vs. 12/1195, 1%; p=0.0001), with increased complex karyotype (14/85, 16% vs. 24/1195, 2%, p<0.001) and enrichment of TET2 (30/85, 35% vs. 203/1195, 17%; p<0.0001) and BCOR (14/85, 17% vs. 96/1195, 8%; p=0.01). Compared with single CMS, double NCMS mutations were less associated with NPM1 (57/236, 24% vs. 681/1195, 57%; p<0.0001) and enrichment of IDH2 (87/236, 37% vs. 251/1195, 21%; p<0.0001) and ASXL1 (33/236, 14% vs. 72/1195, 6%; p<0.0001). These clinicogenomic differences extended to OS: in pAML, NTD mutations had inferior median OS vs both single CMS mutations (7.8 vs. 16 months, p=0.009) and other single NCMS domains (7.8 vs. 17 months, p=0.008). Double NCMS similarly showed inferior median OS vs single NCMS mutations in both pAML (12 vs. 16 months, p=0.004) and sAML (6.5 vs. 9.9 months, p=0.022). Notably, double NCMS showed OS approaching that of single CMS in pAML (13 vs. 16 months, p=NS), suggesting double NCMS may carry single CMS comparable risk.
Transcriptomic analysis of an independent AML cohort (n=615; 477 wild-type, 74 single CMS, 56 single NCMS, 2 CMS/NCMS, and 6 double NCMS) revealed distinct subgroup-specific transcriptional programs consistent with these clinicogenomic associations. Of 13,124 genes analyzed, 620 were differentially expressed (42↑, 578↓) in single CMS and 379 (20↑, 359↓) in single NCMS DNMT3AMT AML compared with DNMT3AWT, vs only 10 (1↑, 9↓) and 23 (4↑, 19↓) in CMS/NCMS and double NCMS, respectively, consistent with the smaller sample sizes (n=2, n=6). Despite fewer DEGs, double NCMS mutations showed disproportionately larger fold-changes (up to log2FC -23 versus -6 in single mutant groups), indicating more severe dysregulation within a smaller gene set rather than a globally attenuated response. Double NCMS samples showed greater gene expression heterogeneity, with 2 of 6 diverging distinctly, consistent with increasing variability at higher allelic complexity. Notably, CMS/NCMS mutations showed pathway enrichment dominated by myeloid and immune-relevant signatures at higher gene ratios, suggesting higher mutation burden driving more hematopoietic specific transcriptional program.
Conclusions
Together, these findings demonstrate that integrating protein domain localization with allelic complexity identifies biologically distinct DNMT3A mutated subgroups characterized by unique transcriptional programs, co-mutational architecture, and clinical behavior, providing a more comprehensive framework for risk stratification than either classification alone.
DISCLOSURES
Asmita Shukla, MD: No relevant conflicts of interest to declare.