Background Spatial organization of immune cell hubs in tumor microenvironments are increasingly linked to response to therapy across diverse tissue types. Bone marrow (BM) provides a rich context for investigating cellular hubs in relapsed acute myeloid leukemia (AML). Studying the AML BM after hematopoietic stem cell transplant (HSCT), may provide new insights into the mechanisms of anti-leukemic immunity in patients treated with adoptive cellular therapy (donor lymphocyte infusion [DLI]). A recognized challenge is that conventional processing of clinical bone marrow samples involves formalin fixation and paraffin embedding (FFPE) and decalcification, degrading RNA and limiting transcriptional profiling. Additionally, computational frameworks are needed to define and compare spatial neighborhoods between responder (R) and nonresponders (NR) patient specimens.

Methods We applied CO-Detection by indEXing (CODEX),¹ a protein-based approach for spatial analysis, to assess the spatial relationships of immune cells on 7 pre-DLI (4 R, 3 NR) and 5 post-DLI (2 R, 3 NR) bone marrow core biopsy FFPE specimens. Images were segmented with DeepCell.² We identified shared spatial colocalization patterns, defined as ‘niche types’, across samples by clustering cell types according to the cellular composition of their neighborhoods. Altogether, we identified 37 distinct niche types in Rs and 19 distinct niche types in NRs.

Results The R niche types were commonly composed of combinations of diverse immune cell populations, including B, T, myeloid, and HSC/leukemia cells across the course of therapy, whereas those identified in NRs lost this diversity post-DLI, as quantified by Shannon entropy (figure 1A, B). Additionally, we observed higher abundance of T and B cells dominating niche types enriched in Rs vs NRs pre-DLI (p=0.004 for T cells, p=0.04 for B cells), whereas myeloid cells exhibited higher proportions in NRs vs Rs pre-DLI (p=0.002).

Deep annotation of T cell phenotypes revealed a shift in T cell state from exhausted pre-DLI (i.e. expressing TIM3, CTLA-4) to effector post-DLI (i.e. expressing CD57, OX-40) (figure 1C, E-F). Post-DLI niches enriched in Rs had higher expression of CD57 and OX40 on T cells (p<1x10^-15), while NRs displayed enrichment of niches expressing TIGIT post-DLI (figure 1D, G). This striking transition to effector memory and cytotoxic T cell states in the R BM complemented our temporal scRNA-seq analysis.^{3 4}

Conclusions Our findings highlight BM niche diversity and less exhausted phenotype of T cells to be key determinants of effective anti-leukemic response.

Acknowledgements We thank Doreen Hearsey and all staff from the Ted and Eileen Pasquarello Tissue Bank in Hematologic Malignancies for excellent technical support with banking of clinical samples. We thank patients for their generous contribution of research samples for this study. We are thankful to Dana Pe’er, David Knowles, Nicolas Beltran-Velez, Siyu He, Lingting Shi, Michael Pressler and Mingxuan Zhang for helpful discussions and feedback. We acknowledge funding from NIH NCI grant R00CA230195 and R01HG012875, Lavine Family Foundation, Lubin Family Foundation, Leukemia and Lymphoma Society grant SCOR-22937-22, and Columbia University Kaganov Fellowship.

References

1. Goltsev Y, Samusik N, Kennedy-Darling J, Bhate S, Hale M, Vazquez G, et al. Deep profiling of mouse splenic architecture with CODEX multiplexed imaging. Cell. 2018;174:968–981.e15.

2. Greenwald NF, Miller G, Moen E, Kong A, Kagel A, Dougherty T, et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning. Nat Biotechnol. 2022;40:555–565.

3. Park C, Mani S, Beltran-Velez N, Maurer K, Gohil S, Li S, et al. DIISCO: A Bayesian framework for inferring dynamic intercellular interactions from time-series single-cell data. bioRxiv. 2023. doi:10.1101/2023.11.14.566956.

4. Maurer K, Park CY, Mani S, Borji M, Penter L, Jin Y, et al. Coordinated immune cell networks in the bone marrow microenvironment define the graft versus leukemia response with adoptive cellular therapy. bioRxiv. 2024. doi:10.1101/2024.02.09.579677.

Ethics Approval The study was approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board. All research was conducted in accordance with the Declaration of Helsinki.

• Download figure
• Open in new tab
• Download powerpoint

Abstract 1232 Figure 1

CODEX spatial mapping of coordinated immune cell types in BM core biopsies. (A,B) Niche types identified in Rs (A) and NRs (B). (C,D) Expression of CODEX markers for T cells in each niche. (E-G) Example windows of niches and selected markers