Introduction: Although myelodysplastic neoplasms/myelodysplastic syndromes (MDS) are driven by genetic mutations, their diagnosis relies on morphologic evaluation of abnormal hematopoiesis in the bone marrow. Only a small number of genetic abnormalities define bone marrow morphologic features in MDS, such as those harboring SF3B1 mutations and deletions of chromosome 5q. We hypothesized that there are additional genetic alterations and dysplastic morphologic features in MDS that have not been well-described in the literature. We assessed genetic morphologic associations between 50 commonly mutated genes and 10 morphologic features in a cohort of MDS bone marrows with a high degree of dysplasia. We validated and extended these findings in an independent cohort of myeloid neoplasms. STAG2 mutations were strongly associated with specific dysplastic megakaryocytic and myeloid cell features in both cohorts. Methods: We reviewed bone marrow core biopsies and aspirates from 272 MDS clinical cases (2013-2014) and selected 89 for further study based on the presence of prominent dysplastic features. Each case was independently evaluated by two pathologists for the following dysplastic features: small hypolobated megakaryocytes, widely separated megakaryocyte nuclei; abnormal myeloid nuclear segmentation, myeloid hypogranulation, Auer rods; ring sideroblasts, erythroid binucleation, irregular erythrocyte nuclei, erythroid nuclear cytoplasmic dyssynchrony, and erythroid karyorrhexis. Targeted next generation sequencing was performed using a 50 gene panel comprising the most common somatic myeloid mutations. For each dysplastic feature, univariate risk analysis was used to identify genetic mutations for use in multivariable analysis. Significant associations were validated through blinded morphologic review in an independent cohort of 155 bone marrow biopsies of myeloid neoplasms, comprised of 83 myelodysplastic syndromes, 33 myelodysplastic/myeloproliferative neoplasms and 39 acute myeloid leukemias, diagnosed between 2014 and 2019 and selected for mutations identified in the initial discovery cohort ( STAG2, RUNX1, SRSF2, ASXL1 and SETBP1) . Results: Pairwise analyses demonstrated broad patterns of genetic associations with myeloid/megakaryocytic dysplasia versus erythroid dysplasia (Panel 1). In multivariable analysis, STAG2 and SRSF2 mutations were significantly associated with megakaryocytes with widely separated nuclei ( STAG2, OR = 25.5, 95% CI [4.17, 493.5], P = 0.003; SRSF2, OR = 10.1, 95% CI [2.07, 75.0], P = 0.008). Mutations in STAG2 and SETBP1 were significantly associated with abnormal myeloid nuclear segmentation ( STAG2, OR = 7.08, 95% CI [1.94, 28.0], P = 0.003; SETBP1, OR = 12.2, 95% CI [1.29, 267.4], P = 0.04). Furthermore, STAG2 mutations were significantly associated with myeloid cell hypogranulation (OR = 12.7, 95% CI [3.10, 86.3], P = 0.002). Because this initial cohort was relatively small, which led to wide confidence intervals suggesting estimation instability, we validated and extended these findings in an independent cohort of myeloid neoplasms. In the validation cohort (enriched for mutations in STAG2, SRSF2, and SETBP1), STAG2 mutations were significantly associated with separated megakaryocyte nuclei (univariate OR = 4.18, 95% CI [1.98, 8.80], P = 0.0002), abnormal myeloid nuclear segmentation (univariate OR = 3.47, 95% CI [1.47, 8.22], P = 0.008), and myeloid cell hypogranulation (multivariable OR = 5.28, 95% CI [2.18, 13.4], P = 0.0003) (Panel 2). In contrast, morphologic associations with SRSF2 and SETBP1 mutations were not validated. Conclusion: STAG2 genetic alterations are strongly associated with the presence of separated megakaryocyte nuclei, abnormal myeloid nuclear segmentation and myeloid cell hypogranulation in MDS. Our data suggest that STAG2 may affect pathways involved in nuclear shape control and secondary granule formation. Panel 1: Pairwise associations between gene mutation and bone marrow morphology in MDS. N, number. Panel 2: Multivariable (black) and univariate (white) analysis in two independent cohorts showing significant gene associations with megakaryocytes with separated nuclei, abnormal myeloid nuclear segmentation, and myeloid cell hypogranulation. CI, confidence interval.