Spatial transcriptomics analysis identifies a tumor-promoting function of the meningeal stroma in melanoma leptomeningeal disease

Abstract

Leptomeningeal disease (LMD) remains a rapidly lethal complication for late-stage melanoma patients. Here, we characterize the tumor microenvironment of LMD and patient-matched extra-cranial metastases using spatial transcriptomics in a small number of clinical specimens (nine tissues from two patients) with extensive in vitro and in vivo validation. The spatial landscape of melanoma LMD is characterized by a lack of immune infiltration and instead exhibits a higher level of stromal involvement. The tumor-stroma interactions at the leptomeninges activate tumor-promoting signaling, mediated through upregulation of SERPINA3. The meningeal stroma is required for melanoma cells to survive in the cerebrospinal fluid (CSF) and promotes MAPK inhibitor resistance. Knocking down SERPINA3 or inhibiting the downstream IGR1R/PI3K/AKT axis results in tumor cell death and re-sensitization to MAPK-targeting therapy. Our data provide a spatial atlas of melanoma LMD, identify the tumor-promoting role of meningeal stroma, and demonstrate a mechanism for overcoming microenvironment-mediated drug resistance in LMD.

Keywords: CSF; MAPK therapy; brain; central nervous system metastasis; drug resistance; leptomeningeal disease; leptomeninges; melanoma; metastasis; pia.

Graphical abstract

Figure 1

Spatial atlas of melanoma LMD compared with other sites of disease (A) Schematic of the spatial transcriptomic analysis workflow, including tissue selection, preparation, sequencing, data quality control, and cell-type deconvolution analysis. Created with BioRender.com. (B) Tissue maps showing spatial cell-type deconvolution for each spot. The cell-type color key is the same for (B) and (C). (C) Bar graphs showing the average deconvolved cell proportion for the major cell types identified in each sample, organized by tissue of origin. (D) Scatterplots showing the mean proportion of spots predominantly comprised of immune and stromal cells in extra-cranial versus leptomeningeal metastasis samples. Error bars indicate standard error.

Figure 2

Spatially driven signaling in melanoma LMD (A) Schematic showing the basic principles behind the algorithm used to identify KEGG pathway enrichment that have significant spatial patterns (left) and a heatmap highlighting the quantity and similarities/differences of KEGG pathways showing significant spatial pattern within each tissue sample (right). (B) Spatial tissue maps showing the position of the stroma, immune, and tumor spots along with tissue maps visualizing the average gene set expression for major pathways important in melanoma biology and drug resistance for each LMD sample. (C) A schematic showing the workflow for determining which genes show a correlation between the level of expression and the distance between tumor and stroma spots (see STAR Methods, Gene expression gradients at tissue niche interfaces, for details). (D) A heatmap of the genes showing a significant correlation between expression and the distance between tumor and stroma spots across multiple LMD samples. (E) Spatial tissue maps showing the position of the stroma, immune, and tumor spots along with tissue maps visualizing the log2 gene expression for GFAP and SERPINA3, and the log2 max expression for sets of genes that are MAPK targets, mTOR targets, and CAF markers (Table S4).

Figure 3

Meningeal cells are required for melanoma survival in the CSF environment (A) Representative microscopy images showing GFP-tagged WM164 melanoma cells treated with 3 μM vemurafenib (BRAFi) or DMSO control in monoculture versus co-culture with primary meningeal cells (left). Bar graphs show the quantification of GFP+ cells across 10 biological replicates relative to the respective vehicle-treated controls (right). Scale bars represent 400 nm. (B) MTT assay showing the relative growth of WM164 melanoma cells treated with increasing doses of vemurafenib (BRAFi) in control medium versus medium conditioned by the co-culture of melanoma with primary meningeal cells. Data represent four independent biological replicates. (C) Representative microscopy images showing GFP-tagged WM164 melanoma cells treated with 3 μM vemurafenib (BRAFi) or DMSO control in monoculture versus co-culture with primary meningeal cells in the context of CSF. Scale bars represent 400 nm. (D) Quantification of the GFP+ WM164 melanoma cells treated with 3 μM vemurafenib (BRAFi) or DMSO control in monoculture versus co-culture with primary meningeal cells for 10 days in CSF (left). WM164 melanoma cells treated with 3 μM vemurafenib (BRAFi) or DMSO control in with regular CSF vs. CSF conditioned by co-culture of melanoma with primary meningeal cells over 13 days (right). Data representative of three biological replicates. Significance was calculated between the co-culture or conditioned medium arm and the respective monoculture control arm. (E) Flow cytometry assessment of cell cycle using PI staining in WM164 melanoma cells treated with 3 μM vemurafenib or DMSO in the context of direct co-culture with primary meningeal cells or with CSF conditioned by co-culture of melanoma with primary meningeal cells. (F) Western blot analysis of WM164 melanoma cells treated with 3 μM vemurafenib or DMSO control in fresh CSF vs. conditioned CSF showing abundance of SERPINA3, pERK(Thr202/Tyr204), ERK, pAKT (Ser473), and AKT. Data represent three independent biological replicates. (G) Western blot analysis of primary meningeal cells stimulated with WM164-conditioned CSF showing the abundance of SERPINA3, cyclin D1, TGF-β1, TGF-βR, and fibronectin. SERPINA3, Cyclin D1 and Fibronectin were blotted on the same membrane and therefore have same loading control. TGF-β1 and TGF-βR were blotted on the same membrane and therefore have same loading control. Data represent three independent biological replicates. All statistical significance was assessed using Student’s t test ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Error bars represent standard error.

Figure 4

Meningeal cells promote growth and resistance to MAPKi in vivo (A) A schematic outlining the experimental cohorts of the animal experiment. Mice were injected subcutaneously with 1 million WM164 cells alone, 1 million WM164 cells with 1 million primary meningeal cells (MCs), or 1 million WM164 cells with 1 million dermal fibroblast (DF) cells. Mice were fed a control rodent diet (control) or a rodent diet formulated with 200 mg/kg dabrafenib and 2 mg/kg trametinib (MAPKi). Created with BioRender.com. (B) Line graph showing the average growth of tumors in each experimental cohort over time, with error bars depicting standard error. Tumor size was fixed following the endpoint of animals until the last animal in the cohort reached the endpoint. (C) Square-root transformation and slope visualization for each experimental cohort in (B). (D) p values for each two-way comparison of slope among experimental cohorts (ANCOVA). Icons created with BioRender.com. (E) Bar graph visualizing the average tumor volume in each experimental cohort on day 21. Data represent mean ± standard error for three biological replicates per cohort. Statistical significance was assessed using Student’s t test ∗p < 0.05, ∗∗p < 0.01. Error bars represent standard error.

Figure 5

Abrogating SERPINA3/IGF1R signaling sensitized melanoma to MAPKi in the CSF environment (A) Representative microscopy images showing GFP-tagged WM164 melanoma cells treated with DMSO control, 100 nM dabrafenib/10 nM trametinib (MAPKi), 1 μM linsitinib (IGF1Ri), or the triple combination in monoculture versus co-culture with primary meningeal cells in the context of CSF. Scale bars represent 400 nm. (B) Quantification of the GFP+ WM164 melanoma cells treated with DMSO control, 100 nM dabrafenib/10 nM trametinib (MAPKi), 1 μM linsitinib (IGF1Ri), or the triple combination in monoculture versus co-culture with primary meningeal cells for 15 days in CSF. Data representative of two biological replicates. (C) Western blots showing knockdown of SERPINA3 with two individual SERPINA3 siRNAs (siRNA1 and siRNA2) in the WM164 melanoma cells and the HMC meningeal cells. Scrambled sequence siRNA was used as control (Con siRNA). (D) Representative microscopy images showing GFP-tagged WM164 melanoma cells treated with 3 μM vemurafenib (BRAFi) or DMSO (control) in monoculture versus co-culture (Co) with primary meningeal cells in the context of CSF following knockdown of SERPINA3 (siRNA1 and siRNA2) or control siRNA (Con siRNA) in both cell types. Scale bars represent 400 nm. (E) Quantification of GFP-tagged WM164 cells from (D). Data represent two biological replicates. Statistical significance was assessed using Student’s t tests (B, C, and E). Significance is denoted as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant. Error bars represent standard error.