Immune phenotyping of MDS bone marrow uncovers disease independent immune signature predictive of lenalidomide response
Lenalidomide is currently the standard of care in del(5q) myelodysplastic neoplasms (MDS), but has shown effectivity in a subgroup of non-del(5q) lower-risk MDS-patients (LR-MDS). The presence of del(5q) is the strongest and most consistent predictor of erythroid response, whereas in non-del(5q) MDS additional factors such as higher progenitor B-cell percentages, absence of ring sideroblasts, and low numbers of somatic mutations have been associated with response. Multivariate analyses consistently highlight immune-related bone marrow features as independent predictors of lenalidomide benefit, supporting their potential utility in refining patient selection beyond cytogenetic classification. This study aimed to identify BM immune signatures and profiles predictive for lenalidomide response in LR-MDS-patients.
-
We conducted an extensive flow cytometric analysis of immune cells in the BM using markers identifying stem cells, lymphoid and myeloid immune cell subsets, and their activation status. The baseline characteristics of the immune profiling cohort were enriched for SF3B1mut and responding non-del(5q) LR-MDS compared to the original H89 cohort. Additionally, we did not observe differences in the type and number of mutations between responders and non-responders nor did we find differences in the variant allele frequency. When we compare all features we found a number of different features to distinct responders and non-responders across different panels. In order to identify cellular programs that distinguish patients responding to lenalidomide, we calculated the correlation between all features in order to define patterns within either responders or non-responders. For responders we found there was a positive correlation of T cells expressing CD39 and activation markers including ICOS and PD1, but also between T cells and dendritic cells expressing CD86, suggesting the presence of tumor-reactive T cells and activated dendritic cells, whereas these correlations were decreased in non-responders. Finally, we aimed to select predictive biomarkers for response utilizing a computational approach for biomarker identification. Using a L1-regularized (LASSO) logistic regression we identified a set of nine immunophenotypic features that are able to distinguish responders from non-responders, irrespective of previous described predictors of lenalidomide response.
Together, these findings provide evidence for a discrete, clinically actionable immune signature that may predict lenalidomide response in MDS. This signature and supports the development of biomarker-guided patient selection strategies, which are currently lacking for non-del(5q) MDS. The results of this study underscore the value of profiling based on immune signatures in an era of precision medicine in MDS.