The dura consists of diverse immune and non-immune cell types

To better understand the cellular composition of human dura, we performed scRNA-seq on samples of human dura and a subset of matched and non-matched primary meningioma samples derived from patients undergoing craniotomy for resection of intracranial meningiomas, which arise from the dura and thus are anatomically attached to this meningeal layer (Additional file 1: Table S1). In surgical resection of meningiomas, if possible, an adjacent region of dura grossly uninvolved with the tumor, as defined by the surgeon, is normally resected to ensure maximal tumor resection and reduce the risk of recurrence [19, 20]. This grossly uninvolved dura, which we define as “non-tumor-associated” dura, was subsequently harvested and used in our analyses. In total, seven dura samples and six primary meningioma samples (four matched and two non-matched) were dissociated and analyzed using scRNA-seq (Fig. 1A, Methods, Additional file 3). We first characterized the non-tumor-associated dura samples, performing unsupervised clustering and uniform manifold approximation and projection (UMAP) [22] analysis on 22,460 cells (Fig. 1B). Cells were initially classified into three cell populations using common markers for endothelial cells (PECAM1, CDH5, KDR), mesenchymal cells (COL1A1, COL1A2, LUM, DCN, ACTA2, RGS5), and immune cells (PTPRC, CD3E, SPI1, CD14) (Fig. 1C and Additional file 1: Table S2). The majority of cells were immune cells (10,423 cells), followed by endothelial cells (6283 cells) and mesenchymal cells (5754 cells). Each patient sample was represented in each of the three major cell types (Fig. 1D). These data demonstrate that the dura harbors a diverse cell population of both immune- and non-immune-derived cell types.

Fig. 1
figure 1

Single-cell preparation and sequencing shows diverse cell landscape of human dura. A Illustration of both non-tumor-associated dura and tumor resection and single-cell library preparation. B UMAP visualization of single-cell RNA-seq data identified by cell type. C Representative gene expression of select cell type gene markers. D UMAP visualization of single-cell RNA-seq data highlighting cells originating from each individual patient sample

Immune cell composition of non-tumor-associated dura

We next focused on resolving the immune cell (PTPRC+, which encodes CD45) landscape. To this end, we performed graph-based clustering and UMAP visualization on 10,423 immune cells (Fig. 2A). Clusters were characterized using a combination of previously reported markers and differentially expressed genes (Fig. 2B, Additional file 1: Table S2, Additional file 4) [29,30,31,32,33,34,35,36,37,38]. This revealed an appreciable population of lymphoid cells, including T cells, NK cells, B cells, and plasma-like B cells, as well as myeloid cells, including monocytes, macrophages, DCs, and mast cells. Expression of important marker genes for lymphoid and myeloid cells were visualized in the UMAP layouts for the indicated cell populations (Fig. 2C, D, respectively). T cells represented the majority of lymphoid cells observed (5458 cells) and included naive/central memory T (TCM) cells (1415 cells; which express SELL, CCR7, LEF1, TCF7, KLF2); CD4+ effector memory T (TEM) cells (1575 cells; SELL-, CCR7-, IL7R); CD8+ TEMs (1281 cells; SELL-, CCR7-, IL7R, CD8A, CD8B, GZMKhi, CXCR3hi); resident memory T (TRM) cells (251 cells; CD69, NR4A2, IL7R); and CD8+ cytotoxic T cells (CTLs) (936 cells; PRF1, NKG7, ZNF683, GZMB, CD8A, CD8B). Other lymphoid cell types identified included natural killer (NK) cells (571 cells; PRF1, NKG7, GZMB, KLRD1, KLRF1, CD3-), B cells (309 cells; CD79A, MS4A1, MHC class II+); and plasma-like B cells (35 cells; IGHG3, IGHA1, DERL1, FKBP11). Meanwhile, the monocyte/macrophage/DC population (3889 cells) was identified by monocyte-related (CD14, VCAN, S100A8, S100A9, MHC class II+), macrophage-related (CD14, RNASE1, C1QA, C1QC, FCAR, GPNMB), and DC-related (CD14-, ITGAX, THBD, and IL3RA) markers. As the cell identities of these clusters were not clearly identifiable by gene marker sets at this resolution, we initially defined this population as a general myeloid compartment composed of monocytes, macrophages, and DCs. Additionally, we identified mast cells (161 cells; GATA2, KIT, HPGDS) in the myeloid compartment. Finally, we sought to compare our data set to a previously published data set [10] that focused on immune cells in murine dura. We selected the top 20 differentially expressed genes (DEGs) for each general cell type in the mouse data set, identified their respective human homologues, and examined their expression in our human data set (Additional file 1: Fig. S1A, Additional file 5). Though not all genes had human homologs nor were expressed in the human data, we found specific expression of murine T/NKT, NK, migDC, and D-BAM markers in the corresponding human immune cell types. These data demonstrate that human dura is made up of a diverse population of immune cells, similar to murine dura, and cell-type-specific gene expression signatures are similar across species.

Fig. 2
figure 2

Immune cell composition of human dura consists of functionally diverse lymphoid and myeloid cell types. A UMAP visualization of immune cells identified by cell type (C1: naive/central memory T (TCM) cells; C5, C6: CD4+ effector memory T (TEM) cells; C2: CD8+ TEM cells; C11: resident memory T (TRM) cells; C3: CD8+ cytotoxic T cells (CTLs)). B Dot plot of gene expression of select cell type gene markers. C, D UMAP visualization of select lymphoid and myeloid marker gene expression. E UMAP visualization of dura DCs (C0, C1, C3, C4, C6, C7: DC-like; C2, C5, C8: migDC-like). F Expression heatmap of top 10 genes of top 10 principal components with hierarchical clustering and associated functional enrichment analysis of gene clusters

To better resolve the monocytes, macrophages, and DCs, we isolated all myeloid cells (excluding mast cells) for further analysis, including dimensionality reduction, clustering, and cell type annotation. We performed graph-based clustering and UMAP visualization on 3889 cells (Additional file 1: Fig. S1B), which revealed one monocyte/macrophage and two DC populations, DC-like and migratory DC-like (migDC-like), identified using marker gene sets (Additional file 1: Table S2, Additional file 1: Fig. S1C). Monocytes and macrophages were grouped together as these cells expressed varying levels of both marker gene sets. The low expression levels of markers previously associated with human microglia, such as AIF1, C1QA, and GPR34 [26], and the expression of markers commonly associated with monocyte-derived macrophages suggested these cells were blood-derived rather than tissue resident. Interestingly, one cluster of cells (C6) was positive for monocyte and macrophage markers but lacked expression of MHC class II genes (HLA-DRA, HLA-DRB1), which suggested that they were myeloid-derived suppressor cells (MDSC-like) [38]. We similarly analyzed the expression of the top 30 DEGs expressed by matched myeloid cell types reported by Van Hove et al. [10] (Additional file 1: Fig. S1D, Additional file 5). Overall, we found that the monocyte/macrophage cluster expressed DEGs from each of the murine myeloid cell types, rather than specific cell types such as classical monocytes or D-BAMs. Notably, the MDSC-like cells under expressed most of these genes. Similarly, the DC-like cluster highly expressed DEGs from each of the murine myeloid cell types though C8 particularly expressed genes differentiating the murine cDC2 cell type. Finally, we found the migDC-like cluster specifically expressed genes identifying murine migDCs. Collectively, these data show the diverse repertoire of myeloid cells within the dura and some conservation across species.

We further characterized the DC population (DC-like: CD14-, ITGAX+ CD1c-, THBDint; migDC-like: CD14- ITGAX- IL3RA+ CCR7+) using unsupervised clustering and UMAP visualization of 2031 cells (Fig. 2E). To investigate variation within these cell populations, we hierarchically clustered the top 10 genes from each of the top 10 principal components (PCs) and used ToppGene [24] to characterize the functional enrichment of co-expressed genes (Fig. 2F, Additional file 2). This highlighted both the cell types present within the tissue, as well as the biological pathways associated with each cell type. DC-like clusters were characterized by the following pathways: “cellular response to interleukin-1,” “positive regulation of cell population proliferation,” and “response to topologically incorrect protein.” Meanwhile, migDC-like clusters were characterized by the following pathway: “cellular response to cytokine stimulus.” Shared pathways included “response to interferon gamma,” “neutrophil chemotaxis,” and “myeloid cell activation involved in immune response.” Interestingly, Chen et al. [33] and Pombo Antunes et al. [36] observed DC clusters similar to migDC-like cells that likewise differentially expressed genes such as CCR7 and LAMP3 and suggested that they are migratory DCs (migDCs) involved in immune cell recruitment. These data suggested that DCs harbored by the dura may be playing a role in establishing the dura immune microenvironment.

Collectively, our analysis demonstrated the presence of both lymphoid and myeloid cell subsets within the dura and underscored the dynamic nature of the dura immune microenvironment.

Endothelial and mesenchymal cells comprise a significant proportion of cells in non-tumor-associated human dura

We next investigated non-immune (PTPRC-) cells by applying graph-based clustering and UMAP visualization followed by cell type annotation. This revealed three main cell types identified by previously reported marker gene sets and differentially expressed genes (Fig. 3A) [40,41,42,43]: endothelial cells (6157 cells; PECAM1, CDH5, KDR, SELE, VWF), fibroblasts (4132 cells; LUM, DCN, COL1A1, COL1A2, COL3A1), and mural cells (1368 cells; ACTA2, MYH11, CNN1, RGS5, PDGFRB, NOTCH3, MCAM, CSPG4) (Fig. 3B, C, Additional file 4). Though C18 (132 cells) was initially a small and indeterminate cluster, as we compared the matched dura and tumor datasets, further analysis revealed that C18 may represent a potential cluster of tumor cells originating from MEN08. Specifically, we observed the presence of copy number variants (CNVs) in 5q and 20q in addition to DEGs, such as COL9A3 and CRABP1, shared with other putative tumor clusters described further below (Additional file 4, Fig. 7). Notably, no other samples harbored detectable populations that could include neoplastic cells. As non-tumor-associated dura is grossly observed to be separate from tumor, the presence of this population could represent an adjacent microscopic cluster of tumor cells. However, due to the limited number of samples and comparatively small number of observed tumor cells in general, we were unable to draw any significant conclusions. We also observed doublets (248 cells), defined by the co-expression of discordant markers for a mixture of cell types, although an increase in the number of transcripts was not detected. The presence of these cell types is consistent with our understanding of the gross structure of dura: a moderately vascularized tissue which harbors a collagen matrix scaffold which underpins the structure.

Fig. 3
figure 3

Functionally diverse non-immune cells comprise a significant proportion of human dura. A UMAP visualization of non-immune cells identified by cell type. B UMAP visualization of select endothelial, mural, and fibroblast markers. C Dot plot of representative gene expression of select cell type gene markers. D UMAP visualization of fibroblast endothelial cells. E Expression heatmap of top 15 genes of top 15 principal components of dura fibroblast cells with hierarchical clustering and associated functional enrichment analysis of gene clusters. F Dot plot of select meningeal fibroblast markers from Desisto et al. [39]. G UMAP visualization of dura endothelial cells. H Expression heatmap of top 15 genes of top 15 principal components of dura endothelial cells with hierarchical clustering and associated functional enrichment analysis of gene clusters. I UMAP visualization of fenestrated endothelium and blood-brain barrier scores

To better characterize the fibroblast cell population, we selected and analyzed these cells using graph-based clustering and UMAP visualization (Fig. 3D). We hierarchically clustered the top 15 genes of the top 15 PCs to identify biological programs (Fig. 3E, Additional file 2). We observed considerable heterogeneity in gene expression profiles and their associated biological pathways with three major hierarchical clusters arising. Cluster 1, which consisted of C0, C2, C4, C8, and C13, was enriched in genes related to “regulation of cell population proliferation,” “regulation of DNA binding,” and “negative regulation of apoptotic process.” These results suggested a distinct subpopulation of proliferating fibroblasts. Another cluster, characterized by C3, C6, C9, and C11, was enriched in genes related to “regulation of cell death” and “cellular response to TGFβ stimulus.” TGFβ stimulus has been shown to induce fibroblast activation and drive scarring in several organ systems [44, 45] and suggested perturbation of the fibroblasts. However, the cause of this activation could be attributed to many sources, such as biological phenomena occurring within the tissue or surgical excision. The third cluster, characterized by C1, C5, and C12, was enriched in genes related to “extracellular matrix organization,” “MHC class II,” and “proteolysis.” MHC class II upregulation has been shown to be induced by IFNγ in dermal fibroblasts [46], and signaling via MHC class II receptors has been shown to lead to cytokine secretion [47]. Based upon these results, these cells may have been actively responding to, and contributing to, immune regulation, in addition to ECM organization. These results highlighted the heterogeneity of fibroblasts in the dura and suggested functional specialization of specific subpopulations.

We next compared the fibroblast clusters to murine meningeal fibroblasts recently described by DeSisto et al. [39] (Fig. 3F). Specifically, we observed high expression of markers associated with murine dura fibroblasts, such as MGP, GJA1, and FXYD5, though not all markers, such as TGFBI, were highly expressed. We found that markers used to delineate different dura layers in mouse models, such as MATN4, CTGF, NPPC, and CRABP2, were not well represented in our data set [48]. Arachnoid markers TAGLN and OGN, also observed in murine dura fibroblasts, were highly expressed in our data set. Interestingly, although C0 and C13 co-clustered with C2, C4, and C8 based upon expression of PC genes (Fig. 3E), only C0 and C13 were enriched in both pia (RDH10) and arachnoid markers (GJB6 and CRABP2). Similarly, C11 was more enriched in arachnoid markers (GJB6 and CRABP2) than C3, C6, and C9 although they were hierarchically clustered together. These results suggested heterogeneity within hierarchical clusters and also that further study may be needed to understand the translation of some meningeal layer-specific murine markers to patient samples. This conclusion was supported by our investigation of the enrichment of murine leptomeningeal fibroblast-like cell (FLC) markers reported by DeSisto et al. [39] based upon data collected by Saunders et al. [49] (Additional file 1: Fig. S2).

We performed a similar analysis on the endothelial cell population by applying graph-based clustering and UMAP visualization (Fig. 3G) and hierarchical clustering of the top 15 genes of the top 15 PCs to identify enrichment of specific biological pathways (Fig. 3H, Additional file 2). Endothelial cells fell into three major hierarchical populations, with cluster 1 (consisting of C0, C1, C3, and C6) distinguished by its expression of genes related to “response to oxidative stress,” “response to abiotic stimulus,” and “cellular response to TGFβ stimulus.” As discussed previously, this may reflect endothelial cell biological phenomena or external stimulus, such as surgical excision or sample processing. Cluster 2, consisting of C2 and C5, were enriched in genes related to “response to oxygen levels,” “response to lipid levels,” “response to interferon gamma,” “collagen-containing ECM,” and “blood vessel morphogenesis.” These results suggested that cluster 2 is metabolically active and involved in vasculature development. Finally, cluster 3, which consists of C4, C7, and C8, exhibits low expression of these genes. As expected, we found that genes related to blood-brain barrier function were not enriched although fenestrated endothelium markers were enriched in hierarchical cluster two (Fig. 3I, Additional file 1: Table S2) [50, 51]. Dural fenestrated vessels have been reported and are important for molecule exchange with the blood [14, 52]. These data demonstrated evidence of a dynamic endothelial cell landscape in the dura layer composed of subpopulations with potentially different functions and the presence of fenestrated endothelium.

Imaging mass cytometry of human dura

Following characterization of human dura at single-cell resolution, we performed imaging mass cytometry on available dura samples, DURA02 and DURA05, to visualize the spatial relationship among these cell types. Similar to our approach with scRNA-seq, we sought to first identify the immune, endothelial, and mesenchymal cell types using specific cell markers (Additional file 1: Table S3). We identified vasculature by the presence of endothelial cells (CD31+, green), which was surrounded by vascular smooth muscle cells (a-SMA+, red) in sample DURA02 (Fig. 4A). Moreover, we observed diffuse presence of collagen (magenta) throughout the tissue as expected. Next, we focused on immune cells, observing concentrated regions of CD45RO expression (cyan) (Fig. 4B). Focusing on T cells, we observed in region 1 the presence of CD8 T cells based upon overlap of CD8 (cyan) and CD3 (magenta). Furthermore, we observed the presence of either naïve, or terminally differentiated, T cells with CD3 and CD45RA overlap. We also observed in region 1 the presence of cells with overlapping expression of CD14 and CD163, which may represent meningeal macrophages (Fig. 4D) [53,54,55]. Iba1, a common microglial marker that is lowly expressed in meningeal macrophages [53,54,55], showed some overlap with CD14 and CD163. However, Iba1+CD14-CD163- cells were likewise observed, suggesting a separate Iba1+ cell population harbored by the dura (Fig. 4D). CD8 T cells were observed to localize nearby CD163+ cells near the vasculature (Fig. 4D). Notably, the majority of these immune cells were observed close to, but outside of, defined CD31+ vasculature. Finally, we observed the presence of GZMB+ CD11b+ cells, often localized within CD31+ vasculature (Additional file 1: Fig. S3A), indicating circulation of cytotoxic immune cells within non-tumor-associated dura. Imaging of DURA05 demonstrated similar results (Fig. 4E–G) with clear CD31+ vasculature surrounded by vascular smooth muscle cells (Fig. 4E) though immune cells, labeled by CD45RO, were sparser and localized around the vasculature (Fig. 4F). Both CD8+ T cells (green arrows) and CD4+ T cells (white arrows) were observed in this sample in region 1 (Fig. 4G). Furthermore, we observed several CD4+ T cells to be adjacent to HLA-DRA+ cells. In contrast, we observed more overlap among CD14, CD163, and Iba1 (Fig. 4H) with fewer Iba1+CD14-CD163- cells. Finally, we observed GZMB+/CD11b+ cells mostly within CD31+ vasculature though infiltration beyond CD31+ vasculature was also noted (Additional file 1: Fig. S3B).

Fig. 4
figure 4

Imaging mass cytometry of a human dura sample reveals intricate spatial relationships among immune, endothelial, and mesenchymal cell types. A, B Imaging mass cytometry of human dura sample DURA02 labeled with markers specified and specific regions of interest (ROIs) highlighted by dashed white boxes. C, D Relative position of each image is denoted by marked number in the top left or right corner. Markers are specified and color coded. White arrows label cells of interest. E, F Imaging mass cytometry of human dura sample DURA05 with markers specified and the specific region of interest (ROI) highlighted by a dashed white box. G, H Relative position of each image is denoted by marked number at the top right corner. Markers are specified and color coded. Green arrows represent CD8+ T cells and orange arrows represent CD4+ T cells. I Hierarchical plot showing the inferred network for MHC-I signaling for immune cells from Fig. 2A. The left and right portions of the plot show autocrine and paracrine signaling to T cells and remaining immune cells, respectively. Solid circles represent the source of MHC class I ligands and open circles represent the target of said MHC class I ligands. Circle sizes are proportional to the number of cells and width of connecting lines indicate the communication probability of said interaction

As cross-presentation of resident macrophages has been suggested [56], and we observed adjacent CD8+ T cells and CD163+CD14+ macrophages, we investigated whether genes associated with such pathways may be overrepresented in the single-cell data. Specifically, applying CellChat [57], which infers and analyzes intercellular signaling pathways, to the immune cell population from Fig. 2A, we identified several signaling pathways that were significantly represented by the single-cell data (Additional file 6). In particular, the monocyte/macrophage/DC population was the most prominent and significant source of MHC-I-related ligands targeting the various T cell populations, with the exception of resident memory T cells (left side of Fig. 4I). Some autocrine signaling was observed with CD4+ TEM cells, CD8+ TEM cells, and CD8+ CTLs. Furthermore, as expected, no significant relationships were observed with non-T cells as the target (right side of Fig. 4I). While these data raised the possibility that interaction between APCs and T cells may occur within the dura tissue itself, further investigation will be required to fully understand the functional roles of these immune cells within the dura as our current conclusions are limited due to the low number of samples and sites of imaging.

Distinct gene expression profiles demarcate immune cells infiltrating meningiomas from those in non-tumor-associated dura

In a subset of patients in our cohort, we were able to collect matched meningioma samples together with non-tumor-associated dura (Additional file 1: Table S1). We analyzed the immune cells of four paired meningioma and non-tumor-associated dura samples composed of 12,581 cells and used the same markers described above to identify cell types. Within each cell type, we observed clear differences in cell state between cells isolated from each location (Fig. 5A, Additional file 1: Fig. S4A, Additional file 1: Table S2, Additional file 4). Notably, we observed that T cells, NK cells, monocytes/macrophages/DCs, and mast cells cluster separately based on tissue origin, whereas B cells from both dura and tumor clustered together. Though dura T cells consist of naïve/TCM cells, TEM cells, and CD8+ CTLs, only TRM cells were observed in the tumor samples.

Fig. 5
figure 5

Non-tumor-associated human dura and meningioma tumor samples show distinctively different immune cell populations. A UMAP visualization of dura and tumor immune cells identified by cell type. B UMAP visualization of dura and tumor monocyte/macrophage/DCs identified by cell type. C UMAP visualization of dura and tumor monocyte/macrophages identified by cell type. D UMAP visualization of select border-associated macrophage (BAM) gene markers and aggregated score. E UMAP visualization of select microglial gene markers and aggregated score. F Violin plots of M1, M2a, M2b, and M2c macrophage polarization state scores of each cluster from Fig. 5C. G Expression heatmap of the top 15 genes of the top 10 principal components with hierarchical clustering and associated functional enrichment analysis of gene clusters. H UMAP visualization of dura and tumor DCs identified by cell type. I Expression heatmap of the top 15 genes of the top 10 principal components with hierarchical clustering and associated functional enrichment analysis of gene clusters

Comparing top DEGs which differentiate dura-originating from tumor-originating T cells, dura T cells’ DEGs were related to T cell migration and function, such as CXCR3 [58] and ITGAL [59], as well as cell motility genes SUSD3 and FGD3 [60, 61] (Additional file 7). In contrast, tumor T cells’ DEGs coded for heat shock proteins, such as HSPA6, HSPA1A, and HSPA1B in addition to genes related to T cell development and function, such as NR4A1 [62, 63] and NR4A2 [62]. Tumor NK cells expressed similar DEGs to tumor T cells and were enriched for genes associated with protein folding and cytokine expression, including genes coding for heat shock proteins HSPA6, HSPA1B, and HSPA1A, and IFNG, a common cytotoxic marker (Additional file 7). However, dura NK cells were enriched for genes that are associated with NK effector function, such as SH2D1B [64] and KLRF1 [65]. Interestingly, NLRC3, a negative regulator of the innate immune response [66], was also overexpressed. Collectively, these data suggested that T cells and NK cells might have different functions in immune regulation depending on tissue of residence. However, as mentioned previously, although both types of tissues were processed similarly, the presence of heat shock proteins may indicate differing responses to the dissociation process rather than reflecting differing cell states within respective tissues. Further investigation will be required to elucidate the clinical implications of these differences.

To further explore the differences among dura and tumor monocytes, macrophages, and DCs, we isolated and reanalyzed both dura and tumor myeloid clusters (excluding mast cells) (Fig. 5B). Marker gene sets were used to differentiate monocyte/macrophages, DC-like, and migDC-like cell clusters (Additional file 1: Table S2, Additional file 1: Fig. S4B). We isolated first dura and tumor monocyte/macrophage clusters and reanalyzed them (Fig. 5C). We used markers associated with microglia and border-associated macrophages (BAMs) in murine models (Additional file 1: Table S2) to determine the potential origin(s) of these tumor monocyte/macrophages [10, 26] (Figs. 5D, E). Both microglial and BAM markers were enriched in tumor-only clusters, and not in dura-only clusters, which suggested that these macrophages were tissue-resident and originated from either the dura or brain parenchyma rather than blood. Similar results were observed via immunohistochemical staining of one matched pair of non-tumor-associated dura and meningioma. Non-tumor-associated dura showed no presence of somatostatin receptor 2 (SSR2), a sensitive marker for meningioma tumor cells [67, 68], low levels of Iba1 (0.228% positive area), CD206 (0.964%), and CD163 (0.368%), and moderate levels of TMEM119 (7.74%), a selective marker for microglia in the brain parenchyma [69] (Additional file 1: Fig. S5). Meanwhile, matched tumor sample showed high levels of SSR2, Iba1 (26.5%), and TMEM119 (39.5%), and moderate levels of CD206 (8.81%) and CD163 (11.9%). Although these results suggested tumor samples contain higher levels of markers associated with BAM and microglia, given the limited number of samples and scope of this study, additional studies will be needed to determine the origin of these macrophages. This limitation of our study was further highlighted by minor discrepancies based on scRNA-seq data, IHC staining, and IMC staining as within non-tumor-associated dura tissue, a considerable population of CD163+ cells and Iba1+ cells were observed via IMC (Figs. 4D, H). Meanwhile, in the IHC staining, we observed low levels of CD206, CD163, and Iba1 expression in non-tumor-associated dura (Additional file 1: Fig. S5). Finally, in the scRNA-seq data, we observed in dura monocyte/macrophages very low levels of MRC1 (which encodes CD206) and heterogenous expression of CD163 and AIF1 (which encodes Iba1), with high expression of these genes in a small population of cells (Fig. 5D, E). Potential reasons for these discrepancies include heterogenous populations of immune cells in non-tumor-associated dura that we were unable to capture with the low number of samples in our data set. Furthermore, as mentioned previously, dura is currently categorized as non-tumor-associated from a gross perspective and the proximity of the non-tumor-associated dura edge to the dural-based tumor mass may vary from case to case. Finally, a difference in mRNA and protein levels, lack of sensitivity to detect rare transcripts, and alterations in cell state may be contributing factors. As we further discuss, additional studies will be required to rigorously define non-tumor-associated dura both from an anatomical and cellular assessment. We then characterized the potential functionality of these macrophages by first assessing the macrophage polarization states in both dura and tumor clusters. Previously reported markers were aggregated to generate scores for M1, M2a, M2b, and M2c polarization [25] (Fig. 5F, Additional file 1: Table S4). Overall, we observed similar M2c scores between dura and tumor clusters. Interestingly, tumor clusters demonstrated marked elevation of both M1 score, which is associated with a pro-inflammatory immune environment, and M2a and M2b scores, which are associated with an anti-inflammatory immune environment [25]. Finally, we hierarchically clustered the top 15 genes of the top 10 PCs to infer biological pathways and determine whether they were distinct based upon tissue source (Fig. 5I, Additional file 2). Overall, we observed a difference in gene expression between dura monocyte/macrophages and tumor monocyte/macrophages with respect to biological pathway enrichment. Most prominently, all dura clusters were enriched for genes associated with “response to sterol” and “myeloid leukocyte migration,” with some clusters enriched in genes associated with “inflammatory response” and “negative regulation of DC differentiation.” Meanwhile, all tumor clusters were enriched for genes associated with “defense response,” “cellular response to IFNγ,” and “response to unfolded protein.” In particular, MHC class II genes such as HLA-DRA and HLA-DRB1 were also upregulated by tumor-specific clusters. Overall, these results indicated that monocyte/macrophages in the dura may have different origins and functional profiles compared to those found in the tumor site.

Both dura and tumor DCs were similarly separated and reanalyzed (Fig. 5H, Additional file 1: Table S2). We analyzed the top 15 genes of the top 10 PCs to infer biological pathways, which revealed considerable heterogeneity (Fig. 5H, Additional file 2). Notably, all dura DC clusters, both migDC-like and general DC-like, were enriched in genes related to “myeloid leukocyte migration.” General dura DC-like clusters (C0 and C8) were enriched in genes related to “cellular response to IFNγ” while migDC-like clusters (C4 and C9) were enriched in genes related to “cell activation,” such as CCR7 and IL2RA, and “regulation of immune system process.” Tumor-specific DC-like cluster C7 was enriched in genes related to “negative regulation of endopeptidase activity” and “inflammatory response” while C3 was enriched in genes related to “cellular response to cadmium ion” and “defense response.” Clusters C1, C2, C5, and C6 were enriched in genes related to “response to other organism” and “regulation of immune system process.” Dura and tumor DCs also have distinct expression profiles, with dura DCs enriched for genes related to cell migration and cell activation and tumor DCs enriched for genes that help mount an immune response.

Given that several studies have demonstrated a functional role for the meninges in CNS immunosurveillance in murine models [14, 70, 71], these comparisons between tumor- and dura-derived immune cells may reflect a similar role for the dura in human disease. However, given that meningioma arises from the meninges itself, another explanation may be that the dura is simply the tissue site through which immune cells migrate to the tumor. Further studies, especially of dura collected from patients with intraparenchymal tumors, will be required to better understand the role of the meninges in response to disease in humans.

TCR analysis of human dura and meningioma samples

To understand T cell clonotypic diversity within both matched dura and meningioma samples, we performed single-cell sequencing on V(D) J region enriched libraries from four dura samples and two matched meningioma samples (Additional file 1: Table S1). We first analyzed the relative frequency of T cell receptors (TCRs) by segregating the predominant clonotype (clone 1) from the rest, which were grouped based upon absolute count (i.e., clones 2–5, clones 6–20, clones 21–100, and clones 101–1000) (Fig. 6A). We observed a greater expansion of the top 20 clonotypes in the dura samples relative to those in the meningioma samples.

Fig. 6
figure 6

TCR frequency and expression overlap in paired non-tumor-associated dura and meningioma samples. A Clonal frequency of the dominant TCR, designated by absolute count, and groupings of TCRs ranked by absolute count. B UMAP visualization of dura and tumor T cells identified by cell type. C, D Alluvial plot demonstrating overlap of the top 15 and 16, respectively, TCRs in paired dura and meningioma samples ranked by relative frequency

Following unsupervised clustering and UMAP analysis of SAMPLE08 and SAMPLE13 T cells alone from Fig. 5A, we observed a clear segregation of dura T cells from tumor T cells (Fig. 6B). Moreover, we generated alluvial plots of the top 15/16 TCRs, ranked by relative frequency with respect to each sample and represented by a distinct color, among the two paired dura and meningioma samples to determine the overlap of TCR presence (Fig. 6C, D, respectively). Strikingly, in the DURA08/MEN08 pair, all top 15 TCRs were identified at varying levels of expansion in both dura and tumor (Fig. 6C). Furthermore, DURA08 and MEN08 had a Morisita index, a measurement of the overlap between two data sets, of 0.484 when comparing the entire TCR repertoires of both samples, indicating a considerable amount of TCR overlap between the two samples (Additional file 1: Fig. S6). In the DURA13/MEN13 pair, 9 of the top 16 most frequently expressed TCRs were present in both DURA13 and MEN13 samples (Fig. 6D). DURA13 and MEN13 have a Morisita index of 0.13, indicating a smaller, but non-zero, overlap of all TCRs compared to the MEN08 and DURA08 pair (Additional file 1: Fig. S6). These data illustrated the T cell clonotypic diversity within the meninges and matched meningioma samples and reveal that TCR clonotypes can be present within both meningiomas and nearby, non-tumor-associated dura sites.

Single-cell analysis demonstrates CNV heterogeneity in meningioma

In addition to analysis of the immune cells from paired dura and meningioma samples, we also performed copy number variant analysis on dura and meningioma pairs to identify putative tumor cells using the R package CONICSmat [28]. Initially, we identified tumor cells in only one paired sample (SAMPLE09), suggesting that tumor cells may have been selected against by the dissociation conditions necessary for dura processing. Therefore, we sequenced two additional meningioma tumor samples (MEN104 and MEN108) dissociated with the Miltenyi human tumor dissociation kit (Miltenyi Biotec), which involves a shorter disaggregation period (Additional file 1: Table S1, Methods). Together, these tumors represented the three WHO grades (MEN104: grade I, MEN108: grade II, MEN09: grade III). DURA09, MEN09, MEN104, consisting of both CD45+ and CD45− fractions, and MEN108, consisting of both CD45+ and CD45− fractions, samples were analyzed with graph-based clustering and UMAP visualization (Fig. 7A, Additional file 1: Figs. S7A and S7B). Using matched immune cells as a reference for CNV detection, we identified a population of cells in each patient sample harboring several chromosomal abnormalities, which we inferred to be tumor cells (Fig. 7B, Additional file 1: Figs. S7A, S7B, and S8). Specifically, the most prominent copy number variants observed for MEN104 were deletion of 19q and 22q (del(19q, 22q)) and amplification of 7p and 7q (amp(7p, 7q)); for MEN108, del(14q, 19q, 22q) and amp(5p, 8q, 9p, 9q, 11p, 15q); and for MEN09 del(1p, 16q) and amp(1q, 6p, 9q, 19q) (Additional file 1: Fig. S8). The majority of these chromosomal abnormalities are consistent with previous observations in WHO grades I and II meningiomas [72]. In WHO grade III meningiomas, amp(16q) and del(6p) are more frequently reported, although del(16q) and amp(6p) have also been observed. Isolation of the tumor cells reveals three distinct clusters with unique DEGs (Fig. 7B, C, Additional file 4).

Fig. 7
figure 7

Analysis of meningioma cells reveals subclonal tumor populations with varying chromosomal abnormalities. A UMAP visualization of one paired dura and meningioma sample and two additional meningioma samples. B UMAP visualization of predicted tumor cells from MEN104, MEN105, and MEN09 samples. C Top 10 DEGs expressed in >50% cells of each sample-specific tumor cluster. DF UMAP visualization of each individual sample-specific tumor cluster with unsupervised clustering and highlighted by respective CNV group (CG#) identities. G–I Expression heatmaps of the top 10 genes of the top 10 principal components of sample-specific tumor cells (DF) with hierarchical clustering and associated functional enrichment analysis of gene clusters

Following CNV and DEG analysis, we isolated, reanalyzed, and investigated sample-specific meningioma cells to better characterize the CNV heterogeneity at the single-cell level. Unsupervised clustering and UMAP analysis were performed in addition to visualization of specific groups of cells based on their respective CNV profiles (Figs. 7D–F). For the tumor cluster derived from MEN104, three major subclonal populations were observed: CNV group 1 (CG1) which contained del(19q, 22q) and amp(7p, 7q), CG2 which contained amp(7p, 7q) and may represent the founding clone, and “Other” which contained the remaining minor CGs consisting of various combinations of the CNVs (Additional file 1: Fig. S8A). Differential expression analysis was used to characterize the expression signature of each CNV group (Additional file 4). To characterize gene expression heterogeneity in the clustered data (Fig. 7D), we analyzed the expression of the top 10 genes of each of the top 10 PCs (Fig. 7G, Additional file 2). Notably, we observed that clusters associated with CG1 (C0, C1, C4, C5, C7, and C8) were enriched in genes associated with several biological pathways, such as “regulation of cell differentiation,” “response to endogenous stimulus,” “response to nutrient levels,” and “regulation of cell motility,” while clusters associated with CG2 (C2, C3, and C6) were under-enriched. These results suggest CG1 cells may be more metabolically active and may be undergoing differentiation to develop a metastatic state. We also observed a cluster of cells (C9) enriched in genes related to “cell division,” indicating an actively dividing subpopulation. Similar analyses were performed for MEN108 (Figs. 7E, H) and MEN09 (Figs. 7F, I). Investigation of MEN108, unlike with MEN104, revealed several CGs with overlapping clustering patterns as observed with UMAP visualization (Fig. 7E). Analysis of the top PC genes revealed less heterogeneity as compared to MEN104 (Fig. 7H, Additional file 2). Most clusters expressed genes related to “oxidative phosphorylation” and “ECM organization” with cluster-specific expression of “response to cell population proliferation” and “cellular response to cytokine stimulus.” One cluster (C10) was enriched in genes related to “mitotic sister chromatid segregation,” suggesting an actively proliferating subpopulation of cells as observed in MEN104. Finally, analysis of MEN09 revealed one CG that characterized the majority of the cells and two smaller CGs (Fig. 7F). Overall, the majority of clusters showed high expression of genes related to signaling responses such as “type I interferon signaling pathway” and “cellular response to cytokine stimulus,” indicating a response of tumor cells to immune surveillance (Fig. 7I, Additional file 2). Several clusters were enriched for genes related to “angiogenesis,” an important process required for tumor development. Other pathways enriched included “oxidative phosphorylation” and “response to unfolded protein,” both of which suggested stress responses, and “vesicle organization.” As with the other tumor samples, an actively dividing cluster of cells was observed as C8 was enriched in genes related to “nuclear division.” Overall, analysis of these tumor cells, derived from tumors characterized by WHO grades I, II, and III, at single-cell resolution indicated the presence of CNV heterogeneity. These CGs were sometimes associated with particular gene expression patterns and respective functional profiles, as in MEN104. However, this was not ubiquitous as MEN108 and MEN09 did not exhibit CNV-associated gene expression profiles.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Disclaimer:

This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (https://www.biomedcentral.com/)

By admin