Profiling the spatial architecture of multiple myeloma in human bone marrow trephine biopsy specimens with spatial transcriptomics

Abstract

The bone marrow microenvironment is intimately linked to the biology that underpins the development and progression of multiple myeloma. However, the complex cellular and molecular features that form bone marrow niches are poorly defined. Here, we used subcellular spatial transcriptomics to profile the expression of 5001 genes in human bone marrow in the context of multiple myeloma. Using this approach, we explored the plasma cell and stroma ecosystem in bone marrow trephine biopsy specimens (herein referred to as trephines) from 21 individuals, including 7 with premalignant disease and 10 with newly diagnosed multiple myeloma. Using spatial transcriptomics in conjunction with an optimized trephine biobanking methodology, we could resolve major components of the human bone marrow microenvironment and reliably characterize distinct plasma cell populations in samples from healthy, premalignant disease and active myeloma. When plasma cells were visualized in the context of location, we detected spatially restricted subpopulations of plasma cells in 5 of 10 newly diagnosed myeloma trephines. Surprisingly, the composition of hematopoietic and stromal microenvironments varied significantly between newly diagnosed myeloma trephines. Furthermore, these differences in microenvironments were also observed within trephines that had spatially restricted plasma cell subpopulations. Thus, these data are not consistent with the hypothesis that a universal bone marrow microenvironment supports the expansion of malignant plasma cells in myeloma. Instead, we propose that myeloma subpopulations form distinct microenvironments and can vary both between patients and spatial locations.

Graphical abstract

Figure 1.

Outline of sample processing and analysis pipeline for spatial transcriptomics. (A) Workflow of preparing human BM trephines for Xenium in situ analysis. (B) DV200 measurements for BM trephines obtained from diagnostic laboratory (n = 5) and formalin-fixed, paraffin-embedded in-house with (n = 8) and without EDTA-based decalcification (n = 16). Error bars represent mean ± standard error of the mean. ∗∗∗P = .0007; ∗∗∗∗P < .0001 (by 1-way analysis of variance [ANOVA] and Dunnett multiple comparisons test). (C-E) Xenium data of a BM trephine collected from a patient with relapsed myeloma (RM-1) using a human multitissue and cancer panel; uniform manifold approximation and projection (UMAP) of Xenium cells (C); spatial plot (scale bar, 1 mm) of panel C with cell type labels derived from published scRNA-seq data (D); and H&E images (E) of PCs (i), pericytes and endothelial cells (ii), megakaryocytes (iii), and erythrocytes (iv), overlaid with transcript expression for listed genes. Each dot represents a single transcript (scale bar, 10 μm). (F) Pixel-level analysis of a disease-free BM trephine (Ctrl-1) with FICTURE using a reference single-cell gene expression profile to define cell types. Magnification of the white-boxed areas (P1 and P2) focusing on endosteal, vascular, and hematopoietic cell–rich microenvironments. Factors are colored based on cell identity inferred from the reference gene expression profile (scale bars, 1 mm [overview] and 100 μm [magnified areas]). Ba, basophil; Eo, eosinophil; H&E, hematoxylin and eosin; Ma, mast cell; RCA, rolling circle amplification; VSMC, vascular smooth muscle cell.

Figure 2.

High-throughput subcellular spatial transcriptomics of human BM. Xenium data of BM trephines from controls (Ctrl-1 to Ctrl-4) and patients with MGUS (MGUS-1 and MGUS-2), SM (SM-1 to SM-5), and MM (MM-1 to MM-10) using a Xenium Prime 5K human pan-tissue and pathways panel. UMAP visualization (A) and proportion (B) of Xenium cells annotated using published scRNA-seq data. (C) CD138 immunohistochemistry and spatial plots of annotated Xenium cells of SM-1 and MM-3 samples (scale bars, 1 mm [overview] and 10 μm [magnified areas]). Ba, basophil; Eo, eosinophil; Ma, mast cell; Osteo, osteo-lineage cells; VSMC, vascular smooth muscle cell.

Figure 3.

Spatial molecular and cellular map of PCs. (A) UMAP plot of the combined data set of PCs from controls (n = 4), MGUS/SM (n = 7), and MM (n = 10). Colors represent PC clusters. (B) UMAP plot of panel A split into controls, MGUS/SM, and MM data sets. (C) Proportion of PC clusters in each sample. (D) Dot plot showing differentially expressed genes between PC populations in panel A. (E) UMAP and spatial plots for representative control (Ctrl-3), MGUS (MGUS-1 and MGUS-2), SM (SM-1 and SM-5), and all MM (MM-1 to MM-10) samples. Scale bars, 1 mm.

Figure 4.

Spatial mapping of genes associated with translocations detected by diagnostic FISH. (A) H&E images of PCs of representative samples overlaid with XBP1 transcripts (green) to identify PCs and expression of genes commonly upregulated in association with the displayed cytogenetic alteration (scale bar, 10 μm). (B) Samples are grouped by cytogenetic category as determined by diagnostic FISH testing. The size and color intensity of each dot represent the relative expression of selected genes commonly upregulated in association with the specific cytogenetic alteration. H&E, hematoxylin and eosin.

Figure 5.

Spatially resolved neighborhood analysis of stromal cells. (A) UMAP plot of the combined data set of stromal cells from controls (n = 4), MGUS/SM (n = 7), and MM (n = 10). (B) Dot plot showing differentially expressed genes between stromal cell populations in panel A. (C) UMAP plot of panel A split into controls, MGUS/SM, and MM data sets. (D) Proportion of stromal cell clusters in each sample. (E) UMAP and spatial plots for representative control (Ctrl-3), MGUS (MGUS-1 and MGUS-2), SM (SM-1 and SM-5), and all MM (MM-1 to MM-10) samples. Scale bars, 1 mm. AEC, arterial endothelial cell; SEC, sinusoidal endothelial cell; VSMC, vascular smooth muscle cells.

Figure 6.

Spatial neighborhood analysis by scider. (A) UMAP and spatial plots of neighborhoods for representative control (Ctrl-3), MGUS (MGUS-1 and MGUS-2), SM (SM-1 and SM-5), and all MM (MM-1 to MM-10) samples. Colors represent transcriptionally distinct neighborhoods. These neighborhoods are deconvoluted with cell labels derived from published scRNA-seq data to generate density heat map for major BM cell types. The heat map shows the expected number of cells per hexagonal bin (scale bar, 1 mm). (B) Proportion of annotated cells in distinct neighborhoods for samples in panel A. Ba, basophil; Eo, eosinophil; Ma, mast cell; Osteo, osteo-lineage cells; VSMC, vascular smooth muscle cells.