Multiple Nf1 Schwann cell populations reprogram the plexiform neurofibroma tumor microenvironment

Abstract

To define alterations early in tumor formation, we studied nerve tumors in neurofibromatosis 1 (NF1), a tumor predisposition syndrome. Affected individuals develop neurofibromas, benign tumors driven by NF1 loss in Schwann cells (SCs). By comparing normal nerve cells to plexiform neurofibroma (PN) cells using single-cell and bulk RNA sequencing, we identified changes in 5 SC populations, including a de novo SC progenitor-like (SCP-like) population. Long after Nf1 loss, SC populations developed PN-specific expression of Dcn, Postn, and Cd74, with sustained expression of the injury response gene Postn and showed dramatic expansion of immune and stromal cell populations; in corresponding human PNs, the immune and stromal cells comprised 90% of cells. Comparisons between injury-related and tumor monocytes/macrophages support early monocyte recruitment and aberrant macrophage differentiation. Cross-species analysis verified each SC population and unique conserved patterns of predicted cell-cell communication in each SC population. This analysis identified PROS1-AXL, FGF-FGFR, and MIF-CD74 and its effector pathway NF-κB as deregulated in NF1 SC populations, including SCP-like cells predicted to influence other types of SCs, stromal cells, and/or immune cells in mouse and human. These findings highlight remarkable changes in multiple types of SCs and identify therapeutic targets for PN.

Keywords: Cell Biology; Expression profiling; Genetic diseases; NF-kappaB; Oncology.

Figure 1. Immune and stromal cell increases in PNs occur long after SC loss of Nf1.

(A) Schematic of the analysis of control and PN cells by scRNA-seq showing analyzed cell numbers. (B) UMAPs of PNs (7-month Dhh-Cre;Nf1^fl/fl), pretumor (2-month Dhh-Cre;Nf1^fl/fl), and corresponding 7-month and 2-month Nf1^fl/fl littermate controls, generated using Seurat. (C) Cell type frequencies across sample types showing all cell types (left); immune cells (middle), and stromal cells (right), showing enrichment of immune and stromal cell clusters in PNs. UMAP cluster numbers (from B) are shown in parentheses for each annotated cell type. (D) Left: Heatmap showing scRNA-seq gene fold changes in 7-month PN cells versus controls, generated using cellHarmony. Right: Differential expression of the same genes in bulk RNA-seq, analyzed as fold changes in 7-month PNs versus controls. The changes were independent of PN growth rate (fast/slow). Broad patterns of differential gene expression across cell types and cell type specific changes are evident.

Figure 2. Normal SC clusters change abundance and gene expression in PNs, and PNs contain SCP-like cells.

(A) Cell type frequencies across SC UMAP clusters. For each cell type, a UMAP cluster number (from B) is given in parentheses. (B) Heatmap showing the top 5 markers for each cluster, generated in Seurat. Shared markers are shown once. For visualization, each cluster was subsampled to 100 cells during heatmap generation. (C) Bar graph shows the percentage of SCs in each cluster that are normal (orange) or show changed gene expression in PNs (green or yellow). (D) A UMAP of NMSCs shows the shift in gene expression in PNs revealed by subcluster analysis. (E) Dot plot showing the top 5 markers for NMSC subclusters. On the y axis, “Identity” numbers represent cell subcluster numbers. (F) A UMAP of SGCs shows the shift in gene expression in PNs revealed by sucluster analysis; 1 SCG appears transiently (white arrows), the other (cluster 1) enlarges in PNs (black arrows). (G) Dot plot showing the top 5 markers for each of 3 SCG subclusters (green and yellow in F).

Figure 3. Enriched HALLMARK gene sets in 7-month-old tumor compared with 2-month-old pretumor SC clusters.

(A) Enrichment analyses using up- and downregulated genes and MsigDB HALLMARK gene sets. (B and C) The EMT genes’ expression patterns in pretumor and tumor SC clusters. Differentially expressed genes were chosen by applying |fold change| > 1.5 and P < 0.05 filters.

Figure 4. Network analysis predicting NF-κB–deregulated PNs confirmed by histological and in vitro analyses.

(A and B) Differentially expressed gene network plots (cellHarmony) showing the central hubs in the NMSC PN cluster (A) and SGCs (B), containing NF-κB transcription factors. Genes shown adjacent to red dots are upregulated, and those next to blue dots are downregulated in 7-month PNs versus 7-month control. (C and D) Immunostaining of tissue sections shows the NF-κB protein p65 (red; C) in mouse PN SCs (expressing EGFP; green) and activated, phosphorylated, p65 (red; D) in mouse PN SC nuclei (green). Yellow arrows indicate colocalization. Scale bars: 15 μm. (E) Normal human SCs (NHSCs) express less p60 and p50 than sphere-forming cells from human PNs (human sphere). In mouse, SCPs from embryonic DRGs contain less p60 and p50 compared with either SCP-like cells from PNs or mouse Nf1^–/– SCPs. Lamin B1 was used as a loading control. See complete unedited blots in the supplemental material. (F and G) Numbers of mouse Nf1^–/– embryonic SCP spheres (F) and human SCP-like cells (G) are slightly reduced by infection with a dominant negative NF-κB (IκB-SS). *P < 0.05 by Welch’s t test. (H) Western blot confirming downregulation of p65 by IκB-SS. (I) p65 immunoreactivity in many cells in human PN tissue sections (1:400). Inset: At 1:100, 38% of cells show immunoreactivity.

Figure 5. Shared cell types between GEMM and human PNs.

(A) Schematic of experimental design for single-cell profiling of human PNs and label transfer to enable comparison to mouse PNs. (B) UMAP shows results of human PN analysis, with annotations after mouse label transfer (Seurat). Note the relative paucity of SCs. (C) Cell type frequencies across sample types in B show multiple types of immune (orange) and stromal (blue/gray) cells. SCs (green) and SCP-like cells (purple) are detectable.

Figure 6. CellPhoneDB analysis of predicted major SC-cell interactions in mouse and human PNs.

CIRCOS plots. (A) Immune cells, stromal cells, and SCs produce midkine (MDK) and pleiotrophin (PTN); these ligands act mainly via the phosphatase PTPRZ1 in SCP-like cells. (B) SCP-like cells and myelinating SCs are predicted to produce PROS1, a ligand for the AXL receptor; AXL is present on immune cells, stromal cells, and some SCs. (C) Myelinating SCs are predicted to produce FGF1 and FGF2, affecting largely stromal cells through the receptor FGFR1. Other SC types are also predicted to express FGF1. (D) SCP-like cells are predicted to produce the ligands APP and MIF, affecting immune cells via the CD74 receptor. M-SC, myelinating SC; NM-SC, nonmyelinating SC.

Figure 7. CD74 is expressed by PN SCs, macrophages, and DCs.

(A) Immunostaining reveals that CD74 expression (green) is low or absent in normal mouse sciatic nerve; expression is elevated in some cells in mouse PNs. Nuclei are stained with DAPI (blue). At right, in human PN tissue sections, CD74 (brown) is expressed by many cells; counterstain is purple. (B) CD74 colocalizes with SCs (CNPase, B), macrophages (CD11b, C), and DCs (CD11c, D) in mouse PNs (white arrows denote colabeled cells). Scale bars: 50 μm. In B–D, dot plots quantify the percentage of cells stained/PN. Horizontal bars denote mean ± SEM.

Figure 8. The gene expression pattens of macrophage clusters.

Based on (A) days after nerve injury (48), (B) nerve injury-associated macrophage subtypes (49), and (C) “don’t-eat-me” and “eat-me” (engulfment receptor) signatures (49).