P-071: Multi-omic analysis of the tumor microenvironment reveals novel associations in a clinical trial of atezolizumab ± daratumumab for relapsed/refractory multiple myeloma
2021
Background We conducted a Phase 1b trial of atezolizumab (A, anti-PD-L1) ± daratumumab (D, anti-CD38), which targets myeloma cells and has immunomodulatory activity, on the hypothesis that the combination may alter the tumor microenvironment (TME) to favor T-cell activation in relapsed/refractory multiple myeloma (RRMM) (NCT02431208). We undertook multi-omic analysis of samples collected from the TME to understand the immune milieu and potential correlations with clinical outcome. Methods Bone marrow mononuclear cells (BMMNC) and bone marrow plasma were collected from 4 different cohorts, Cohorts A (A-monotherapy), D1 (1–3 prior lines, D-naive, A+D combination therapy), D2 (3+ prior lines, D-naive, A+ D), and D3 (D-refractory, 3+ lines of therapy, A+D), at baseline and on-treatment. Bulk RNA sequencing (RNAseq) was performed using longitudinal CD138+ and CD138-enriched fractions. Mass cytometry immunophenotyping (39 marker CyTOF panel, Fluidigm) and proteomic profiling (multiplex Immuno-Oncology assay [OLink)]) was performed on BMMNC and BM plasma. For the unbiased integrative analysis, Similarity Network Fusion (SNF) algorithm was applied to preprocessed CyTOF, OLink and RNAseq data. Results Data specific and fused patient (pt) similarity networks were derived by unsupervised SNF algorithm from CyTOF, OLink, CD138– RNAseq, and CD138+ RNAseq features. At baseline, networks built using a single data type yielded distinct patterns of pt similarity. However, the fused network, integrating information from all four layers of data types, separated the subjects into three groups which distinguished the D-refractory (D3) and the D2 cohort from the D-naive A and D1 cohorts. Notably, SNF applied to post-treatment CyTOF, CD138+ and CD138– RNAseq data resulted in three clusters that recapitulated the treatment cohorts. The three responders resolved in Cluster 1, which included pts in cohorts D1 and D2. Cluster 2 included the D-naive pts who were treated with A-monotherapy, and Cluster 3 included the D-refractory pts treated with A+D-combo. Pairwise analysis comparing matched treatment samples to baseline identified key biomarkers that are differentially expressed between subgroups. The “one-versus-all” comparisons using the hallmark gene sets in the CD138– gene expression data layer revealed that Cluster 1 is enriched for the T cell gamma delta gene signature and the T effector 6 gene signature, which has been associated with response to cancer immunotherapy treatment such as in Cohort A. In the CD138 positive gene expression layer, the dendritic cells gene signature was significantly increased in Cluster 3, which may indicate a mechanism of resistance in the D–refractory pts. Conclusions Unsupervised machine learning-based integrative clustering analysis of baseline and on-treatment samples from multiple immunologic data types revealed novel associations with pt selection and outcome in both D-naive and D-refractory RRMM pts treated with A ± D.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
0
Citations
NaN
KQI