Increased presence of reactive microglia in human H3K27M DMG
We previously reported that microglia, wildtype for histone H3, are present in the tumour mass from DMG H3K27M subjects and that in vitro microglia exposed to patient-derived DMG cancer cells acquired a reactive state holding tumour trophic functions [26]. The presence of microglia was further analysed in archival human DMG tissue samples obtained from the BRAIN UK biobank. A total of four DMG cases, with reported H3K27M mutation, were examined, including three biopsies taken at the time of diagnosis and one post-mortem sample. These DMG cases were compared to two age-matched brainstem control cases, in which the individuals had died from non-neurological causes (Suppl. Table 1). Microglial numbers and activation states were assessed using IBA1 (ionized calcium binding adaptor molecule 1, also known as allograft inflammatory factor 1, AIF1) immunofluorescence staining [27, 28]. The paediatric brain DMG tissue samples exhibited an increased presence of microglia as compared to the control pons tissue samples (Fig. 1a, b; Suppl. Fig. 1). Additionally, an increased fluorescence intensity of IBA1 per cell was observed in the DMG cases (Fig. 1a, c; Suppl. Fig. 1). Collectively, our previous findings and these observations support the notion that microglia play an active role in DMG tumour progression rather than being passive bystanders.
Fig. 1: Increased presence of reactive microglia in human H3K27M DMG tumours.The alternative text for this image may have been generated using AI.
a Confocal microscopy imaging of two human DMG tumours (one biopsy and one post-mortem samples), and one age-matched pons control case (post-mortem sample), with immunofluorescence staining for IBA1. DAPI used as nuclear counterstain, scale bars (50 µm). b IMARIS quantification of the numbers of IBA1-positive cells/mm2, and (c), IMARIS quantification of IBA1 fluorescence intensity/cell from the archival tissues described in panel a. Data are mean ± SEM from independently analyzed microglia from 2 DMG (1 biopsy, 1 post-mortem) and 1 age-matched control human tissues. Statistical annotation ***p < 0.001; for indicated comparisons. See Suppl. Fig. 1, for additional two human DMG tumours, and one age-matched pons control case.
Microglia exposed to DMG H3K27M cells exhibit a transcriptomic response associated with extracellular matrix remodelling
To gain further insights into the reactive state acquired by microglia when exposed to DMG cancer cells and how this pro-tumoral microglial phenotype could contribute to DMG progression, their transcriptomic response was investigated. For these experiments, a segregated co-culture system based on soluble factor exchange was used [26, 29, 30]. BV-2 microglia [31] were exposed to primary human cancer cells originating from a 3-year old female diagnosed with DIPG, SF8628, carrying the H3K27M mutation (thereafter referred as SF8628 DMG cells) [32, 33] or to SF188 cells representing pHGG originating from a 8-year old male right frontal lobe glioblastoma which harbour a H3-wildtype (H3-wt) status (referred as SF188 pHGG cells [34, 35]. Microglia exposed to SF8628 DMG cells or SF188 pHGG cells, for 3 and 6 h, were collected, and bulk RNA-seq was performed on three independent biological replicates (Fig. 2; Suppl. Fig. 2). Volcano plots illustrate the most significantly up- and down-regulated differentially expressed genes (DEGs) in microglia exposed to SF188 pHGG cells or SF8628 DMG cells for 3 h (Suppl. Fig. 2a, b) and 6 hours (Fig. 2a, b), as compared to microglia alone (Fold Change > 1.5 or < −1.5; P-value < 0.05). A direct comparison of the transcriptome from microglia exposed to SF8628 DMG cells compared to microglia alone, versus microglia exposed to SF188 pHGG cells compared to microglia alone revealed that microglia exposed to SF8628 DMG cells exhibited a significantly higher expression of genes related to the extracellular matrix (ECM) (Fig. 2c, d; Suppl. Fig. 2c, d). These ECM-related genes included Fn1 (encoding for fibronectin), Col1a1, and Col6a3 (collagens type I alpha 1 chain and type VI alpha 3 chain, respectively) at the 3-hour time point, as well as Col1a2 (collagen type I alpha 2 chain) at the 6-hour time point. Normalized counts for the microglial expression of those ECM-related genes after 3- and 6-hour exposure to SF188 pHGG cells or SF8628 DMG cells and microglia alone (used as control) are presented in Fig. 2e–h; Suppl. Fig. 2e–h. Collectively, these transcriptomic data indicate that the response of microglia, resident immune cells of the brain, to DMG H3K27M cells is distinct from their response to pHGG H3-wt cells.
Fig. 2: Microglia exposed to DMG H3K27M cells and pHGG cells exhibit different transcriptomic responses.The alternative text for this image may have been generated using AI.
a, b Volcano plots illustrating differentially expressed genes (DEGs) based on the log2(fold change) related to negative log10 (P-value) between microglia exposed to SF188 pHGG cells for 6 h and microglia alone (a) or between microglia exposed to SF8628 DMG cells for 6 h and microglia alone (b). c Volcano plots illustrating DEGs between 6-hour SF8628 DMG cells-stimulated microglia compared to microglia alone and 6-hour SF188 pHGG cells-stimulated microglia compared to microglia alone. Blue dots represent significantly downregulated genes with log2(FC) of maximally -1, and red dots significantly upregulated genes with log2(FC) at least 1. Names of the top 20 genes for up- or downregulated genes are depicted. d Heatmap representation of genes found to be statistically differentially expressed (FC > 1.5) between conditions described in (c). Expression data for these genes were extracted from the comparisons of microglia exposed for 6 hours to SF188 pHGG cells compared to microglia alone or SF8628 DMG cells compared to microglia alone. e–h Fn1, Col1a1, Col1a2 and Col6a3 gene expressions as normalized counts, presented as mean ± SEM, in microglia exposed to SF188 pHGG cells for 6 h, microglia exposed to SF8628 DMG cells for 6 h and microglia alone. Data depicted in this figure originate from RNA-seq analysis of 3 independent biological replicates for each group. Statistical annotations *p < 0.05; **p < 0.01; and ***p < 0.001; for indicated comparisons. See Suppl. Fig. 2, for similar analysis on a 3-hour time point.
Bioinformatic analysis revealed distinct Gene Ontology (GO) enrichment patterns for biological processes between microglia exposed to different tumour genetic backgrounds. GO biological process analysis showed that microglia exposed to pHGG cells express genes involved in immune activation and DNA damage repair at the 3-hour time point (Suppl. Fig. 3a). After 6-hour stimulation, the microglial response to pHGG cells shifted toward regulating protein function and assembly, with an increased expression of genes associated with stress responses (Fig. 3a). In contrast, microglia exposed to DMG cells exhibit an enrichment in biological processes related to cellular adhesion, morphogenesis, and regulation of cell movement, reflecting early tissue remodelling at the 3-hour time point (Suppl. Fig. 3a). After 6-hour stimulation, the microglial response to DMG cells shifted toward roles in developmental processes, synaptic signalling, together with engagement in metabolic and signalling transduction pathways (Fig. 3a). Similar results were obtained from Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, revealing pathways linked to immune activation in microglia exposed to pHGG cells and extracellular matrix remodelling in microglia exposed to DMG cells at the 3-hour time point (Suppl. Fig. 3b). At the 6-hour time point, microglia exposed to pHGG cells displayed pathways associated with metabolic reprogramming, suggesting a shift toward supportive functions. Meanwhile, microglia exposed to DMG cells continued to exhibit pathways related to tissue remodelling (Fig. 3b). To further explore these findings, we investigated the expression patterns of genes involved in the “ECM receptor interaction” KEGG pathway, which was identified as a prominent hit in the DMG context at both time points. At 3 hours, the genes Fn1, Col1a1, Col1a2, Col4a2, Col6a3, and Itga3 were expressed (Suppl. Fig. 3c), while at 6 h, Fn1, Col1a1, Col1a2, and Col6a3 remained highly expressed (Fig. 3c). Heatmap representation for this KEGG pathway shows a significantly higher expression of these genes in microglia exposed to DMG cells as compared to microglia exposed to pHGG cells or microglia alone. Transcription factors and their selective target genes are central players in transcriptional regulation and in the acquisition of transcriptional programs controlling cell phenotypes. The transcription factor-target genes interaction database TRRUST (Transcriptional Regulatory Relationships Unravelled by Sentence-based Text mining), which allows for the identification of transcription factors potentially involved in cellular responses based on the expression of their target genes, further revealed that different sets of transcription factors are involved in the acquisition of the microglial reactive states observed upon their stimulation by DMG cells versus pHGG cells (Fig. 3d; Suppl. Fig. 3d). Hence, these transcriptomic data revealed that the responses of microglia to an exposure to pHGG cells and DMG cells are distinctive in term of their gene expression profile and that the microglial reactive states acquired upon stimulation by DMG cells is characterized by a further ECM-related transcripts enrichment.
Fig. 3: Microglia exposed to DMG H3K27M cells exhibit a transcriptomic response associated to extracellular matrix remodelling.The alternative text for this image may have been generated using AI.
a Analysis of enriched Gene Ontology (GO) terms for biological processes showing the top 10 significant terms sorted by -log10(p-value) for both, microglia exposed to SF188 pHGG cells and SF8628 DMG cells at 6-hour time point. Size of circle represent number of genes included in each GO BP term. b KEGG pathways enrichment analysis that allows identification of significantly affected pathways related to upregulated genes in SF188 pHGG cells-stimulated microglia and SF8628 DMG cells-stimulated microglia. The top 10 significant terms sorted by -log10(p-value) are displayed. Size of circles show number of genes included in each term and X-axis represent calculated enrichment. c Heatmap represent genes included in ECM-receptor interaction KEGG pathway displaying expression of genes of interest between experimental conditions. d TRRUST analysis, which allows the identification of potentially involved transcription factors based on the expression of their target genes in SF188 pHGG cells-stimulated microglia and SF8628 DMG cells-stimulated microglia. The top 5 genes sorted by -log10(p-value) are shown. Data depicted in this figure originate from RNA-seq analysis of 3 independent biological replicates for each group. See Suppl. Fig. 3, for similar analysis on a 3-hour time point.
Inhibition of fibronectin derived from microglia reduces DMG H3K27M cell invasiveness
The induction of Fn1 and Col1a1 gene expression in microglia exposed to SF8628 DMG cells, relative to both microglia exposed to SF188 pHGG cells and microglia alone, was confirmed by RT-qPCR analysis (Fig. 4a, b; Suppl. Fig. 4a), validating the upregulation of ECM-related factors identified from bulk RNA-seq (Fig. 2 and Suppl. Fig. 2). Consistent with these transcriptional changes, immunoblot analysis demonstrated a marked increase in FN1 protein levels in microglia exposed to SF8628 DMG cells for 6, 24 and 48 h, as compared those stimulated by SF188 pHGG or microglia alone (Fig. 4c, d). The COL1A1 protein levels showed only modest trend toward upregulation at the 3-hour time point, as assessed by immunoblot (Suppl. Fig. 4b). Therefore, we decided to validate COL1A1 protein expression by immunocytochemistry in microglia itself, exposed to SF188 pHGG cells or SF8628 DMG cells. Analysis of COL1A1 fluorescence intensity in microglia related to their size shows the same trends to mild upregulation at 3 h time point as immunoblot approach (Suppl. Fig. 4c, d).
Fig. 4: Inhibition of fibronectin derived from microglia reduces DMG H3K27M cell invasiveness.The alternative text for this image may have been generated using AI.
a, b RT-qPCR analysis confirm upregulation of Fn1 mRNA expression in microglia exposed to SF8628 DMG cells, as compared to microglia exposed to SF188 pHGG cells or microglia alone (a) at 3- and 6-hour time point, (b) at 24- and 48-hour time point on. c, d Immunoblot analysis and quantification reveal upregulation of FN1 protein expression in SF8628 DMG cells-stimulated microglia, as compared to SF188 pHGG cells-stimulated microglia or microglia alone (c) at 3- and 6-hour time point, (d) at 24- and 48-hour time point. e–g Quantification of the DMG cells invasion capacity (e) SF8628 in presence of microglia and FN1 inhibitors RGDS and Avapritinib, (f) DIPG-XIII and (g) DIPG-XVII in presence of microglia and FN1 inhibitor RGDS. h Quantification of the effect of Fn1 siRNA in microglia on SF8628 invasion capacity. Data are mean ± SEM from 4 (a–f) and 3 (g, h) independent biological replicates. Statistical annotations *p < 0.05; **p < 0.01; and ***p < 0.001; or exact p-value for indicated comparisons. See Suppl. Fig. 4, for Col1a1 mRNA and ICC protein expression validation and control experiments for the invasion assay.
Given that significant induction of FN1 expression was established as a feature of the response of microglia stimulated by DMG cancer cells, not observed in microglia stimulated by pHGG cancer cells, we decided to explore whether inhibiting this ECM component would alter tumour cell invasiveness.
The RGDS (Arg-Gly-Asp-Ser) tetrapeptide is a well-established antagonist of FN1 [36, 37]. This peptide blocks the integrin-binding domain, preventing tumour cell attachment to FN1. Although originally described as a PDGFRA inhibitor, Avapritinib was recently identified in a large-scale drug repurposing screen of 2471 FDA-approved compounds as a potent inhibitor of FN1 [38]. Avapritinib forms a more stable complex with FN1 than any other screened drug and induces substantial alterations in the secondary structure of the extra domain A (EDA) region of FN1, interfering with its function. Of importance for the present study, in vivo validation for the potential benefit of using Avapritinib in the context of DMG tumours was recently established. Indeed, Avapritinib treatment has been shown to significantly reduce DMG tumour growth in animal models of the disease, and a subset of paediatric and young adult PDGFRA-altered HGG patients showed clinical response to Avapritinib [39]. A key point for our investigation is that microglia do not express PDGFRA (based on the Human Protein Atlas, and observed absence of PDFFRA expression in microglia in the human DMG single-cell RNA-seq dataset described below), inferring that observed effects of Avapritinib on microglia cannot be attributed to PDGFRA inhibition but instead are most likely mediated via FN1 inhibition.
A mixture of SF8628 DMG cells and microglia treated with or without the RGDS peptide or Avapritinib embedded in Matrigel were placed on the upper compartment of a transwell migration assay, and the invasion capabilities of the DMG cells assayed. A significant reduction in tumour cell invasion in the presence of any of the FN1 inhibitors, as compared to controls, was observed for SF8628 DMG cells (Fig. 4e). RGDS and Avapritinib treatments were also shown to reduce the invasion capability of the microglia (in the presence of SF8628 DMG cells) but not of the SF8628 DMG cells alone (Suppl. Fig. 4e, h). The effect of RGDS peptide-treated microglia was also assessed on two additional H3K27M carrying DMG primary cell lines, i.e., DIPG-XIII and DIPG-XVII [40]. providing similar results with the exception that RGDS treatment on its own altered the invasion ability of DIPG XVII cells (Fig. 4f, g; Suppl. Fig. 4f, g).To directly assess whether microglial FN1 contributes to both microglial invasion and microglia-mediated promotion of tumour invasion, we silenced Fn1 expression in microglia using a pool of small interfering RNAs (siRNAs). Efficient knockdown of Fn1 significantly reduced the invasive capacity of microglia themselves and markedly diminished their ability to enhance SF8628 DMG cell invasion (Fig. 4h; Suppl. Fig. 4j).
Collectively, these results suggest that microglia-derived ECM components, particularly FN1, shape the tumour microenvironment and promote the invasive behaviour of DMG cells.
Human DMG H3K27M tumours exhibit an increased expression of tumour-associated myeloid cell-derived ECM components
To gain pathophysiological understandings into the potential role for the uncovered microglial ECM-related factors in the DMG H3K27M tumour microenvironment, a meta-analysis was performed comparing the transcriptomic data generated from BV-2 microglia exposed to SF8628 DMG cells at 3- and 6-hour time points with an independent dataset generated from CD45+CD11b+ expressing cells isolated from human DMG tissue samples at the time of early post-mortem autopsy (Suppl. Fig. 5a). CD45+CD11b+ expression in normal paediatric tissue corresponds to the resident microglial population [24]. However, in the context of a brain neoplasm, CD45+CD11b+ expression does not allow the differentiation between resident microglia and peripherally recruited bone-marrow-derived macrophages, and therefore the analysed cell population should be considered as tumour-associated myeloid cells (TAMs, i.e., microglia and macrophages). This human dataset comprises six cases of DMG and three paediatric control samples derived from normal cortical tissue [24]. The comparative analysis of 3,825 DEGs (defined by a fold change > 2 or < −2, irrespective of statistical significance) from TAMs isolated from human DMG biopsies (as compared to control cases) and 1234 DEGs from microglia exposed to SF8628 DMG cells (as compared to microglia alone) at the 3 h time point, or 1254 DEGs at the 6-hour time point, revealed an overlap for 95 DEGs and 88 DEGs when the 3 h and 6 h time point were used, respectively (Suppl. Table 2).
Enrichment analysis for GO biological process terms using the overlapping DEGs found to be common between TAMs isolated from DMG biopsies and microglia exposed for 3 h to SF8628 DMG cells identified FN1, COL1A1, COL1A2, COL4A2, COL5A2, and COL6A1 as candidate genes whose expression is associated with biological processes such as extracellular matrix organisation, regulation of cell-substrate adhesion, glial cell fate determination, cell-matrix adhesion, and plasma membrane regulation (Suppl. Fig. 5a). Corresponding KEGG pathway analysis confirmed a strong link for these shared DEGs to ECM remodelling (Suppl. Fig. 5b). When microglia exposed for 6 hours to SF8628 DMG cells were used for the comparison with TAMs isolated from DMG biopsies, overlapping DEGs included FN1, COL1A1, COL1A2, COL5A2, and ITGA3, with an enrichment in GO biological process terms for ossification, regulation of Wnt signalling, collagen fibril organisation, response to steroid hormones, and cell-matrix adhesion (Fig. 5a). KEGG pathway analysis at this time point revealed similar ECM-related alterations to those observed at the 3-hour time point (Fig. 5b). Hence, these human transcriptomic data further support a role for microglia (alternatively, more broadly TAMs) in driving a remodelling of the extracellular matrix, including the production of FN1, within the DMG tumour microenvironment.
Fig. 5: Tumour-associated myeloid cells exhibit expression of ECM-related genes in human DMG H3K27M tumours.The alternative text for this image may have been generated using AI.
a Chord diagram representing the top 5 shared GO biological process terms between in vitro assayed SF8628 DMG cells-stimulated microglia (6-hour time point) and TAMs from paediatric DMG with highlighted genes involved in extracellular matrix remodulation. b Sankey diagram combined with dot plot representing top 10 significant terms sorted by -log10(p-value) for shared KEGG pathways. Size of circle represents number of genes included in each term, while X-axis represents calculated Odds ratio. c Uniform manifold approximation and projection (UMAP) of 5 DMG patient tumour samples showing FN1 expression. Each dots represents individual cells. FN1 differential expression in microglia versus all other cell types is included in the panel (d), Immunoblot analysis and quantification of FN1 protein expression in iPSC microglia after 6- and 24-hour exposure with SF8628 DMG cells. e Quantification of the SF8628 DMG cell invasion capability in presence of iPSC microglia and FN1 Inhibitors RGDS and Avapritinib. Data in (a, b) originates from mutual comparison of 3 independent in vitro biological replicates of microglia exposed to SF8628 and 5 human DMG biopsies. Data in (c) originates from 5 DMG patient tumour samples. Human iPSC-related data are represented as mean ± SEM from 3 (d, e) independent biological replicates. Statistical annotations *p < 0.05; **p < 0.01; for indicated comparisons. See Suppl. Fig. 5, for similar analysis on a 3-hour time point.
Microglia are the primary FN1-expressing cells in human DMG H3K27M tumours
Multiple cellular components of the tumour microenvironment can contribute to extracellular matrix assembly, and microglia may not be the only source of FN1. To clarify which cell populations express FN1 in human DMG, we re-analysed a publicly available single-cell RNA-seq dataset generated from 19 pHGG, including five DMG H3K27M cases [41].
Restricting the analysis to the five DMG H3K27M cases revealed that FN1 expression is predominantly localized to the myeloid compartment, with microglia representing the statistically major FN1-expressing cell population within DMG tumours. In contrast, all other investigated cell types showed minimal or negligible FN1 transcript levels (Fig. 5c). When extending the analysis to the full cohort of 19 pHGG samples, the myeloid compartment remained the dominant source of FN1-producing cells (Suppl. Fig. 5d).
To strengthen the human relevance of these findings, we conducted additional experiments using human induced pluripotent stem cell (iPSC)–derived microglia, an emerging tool to study microglial functions in health and disease [42]. Following exposure to SF8628 DMG cells, iPSC-derived microglia exhibited a marked increase in FN1 protein expression compared with both untreated iPSC-derived microglia and those stimulated with SF188 pHGG cells, as shown by immunoblot analysis (Fig. 5d). Moreover, treatment of iPSC-derived microglia with the FN1 inhibitors RGDS or Avapritinib significantly reduced their intrinsic invasion capacity and diminished their ability to promote SF8628 DMG cell invasion in Matrigel-based transwell assays (Fig. 5e; Suppl. Fig. 5f).
Together, these findings support the conclusion that human microglia respond to DMG-derived cues by upregulating FN1 and that microglia constitute a principal source of FN1 within the DMG tumour microenvironment.
Validation of ECM component expression in human DMG H3K27M tissue samples
To gain further validation that the expression of COL1A1 and FN1, two of the identified microglial ECM-related factors, are altered in the context of human DMG tumours, their protein expression was analysed by immunofluorescence in the same four DMG archival tissue samples obtained from the BRAIN UK biobank and used for the detection of microglia depicted in Fig. 1 and Suppl. Fig. 1. Compared to control cases, which exhibited no specific staining, DMG biopsies showed increased expression of both FN1 and COL1A1. As controls, tissue samples from two cases without neurological causes of death were included. IBA1 immunofluorescence co-staining was used to detect the microglial/TAM cell population within these archival tissue samples. Compared to control cases, which exhibited limited to no specific staining, DMG biopsies showed an increased expression for both FN1 and COL1A1. Notably, when comparing diagnostic biopsies to postmortem tissue, we observed a greater area of deposit coverage typical with high signal intensity (Fig. 6; Suppl. Fig. 6).
Fig. 6: Human DMG H3K27M tumours exhibit increased FN1 and COL1A1 protein expression.The alternative text for this image may have been generated using AI.
a Confocal microscopy imaging of two human DMG tumours (one biopsy and one post-mortem samples), and one age-matched brainstem control case (post-mortem sample), with immunofluorescence staining for FN1 and IBA1. b Confocal microscopy imaging of the same cases as in panel a, with immunofluorescence staining for COL1A1 and IBA1. DAPI used as nuclear counterstain, scale bars(50 µm). Representative images originate from 2 independent DMG cases (1 biopsy, 1 post-mortem) with aged matched healthy subject (post-mortem). See Suppl. Fig. 6, for additional two human DMG tumours, and one age-matched brainstem control case.
Collectively, the transcriptomic data from both microglia exposed to DMG cancer cells, TAMs isolated from human DMG biopsies, as well as the protein expression analysis for COL1A1 and FN1 in human archival DMG biobank tissues, support the concept of a microglia/TAMs-mediated remodelling of the ECM within the tumour microenvironment that could impact on the progression of DMG and potentially even hold prognostic value.
Validation of ECM component expression and their prognostic value in DMG H3K27M across human cohorts
Given the rarity of paediatric brain neoplasms and the challenges in obtaining precise control cases, we expanded our analysis to publicly available patient cohorts to further strengthen our findings. First, we utilised genomic and transcriptomic data for DMG H3K27M subjects from the Gabriella Miller Kids First Pediatric Research Program (referred as Kids First), initiative focused on identifying genetic contributors to childhood cancer [43]. and transcriptomic data from the Genotype-Tissue Expression (GTEx) Project, that provide a resource for healthy tissue reference [44]. to evaluate the expression of genes encoding microglial markers and ECM components in DMG H3K27M samples as compared to healthy tissue from anatomically relevant regions, including the spinal cord (Cervical C1) and cerebellum, which are adjacent to the midline (Suppl. Fig. 7a). In the DMG H3K27M subjects, a significant increase in the expression of P2RY12 and TMEM119, microglial markers, was observed as compared to healthy controls (Fig. 7a, b) [45, 46]. The increased gene expression for these microglial markers in bulk RNA-seq from DMG H3K27M tumours, further confirmed the observation made with IBA1 immunofluorescence analysis on archival DMG H3K27M tissue samples, that microglial content is increased in these paediatric brain neoplasms (Fig. 1). In the DMG H3K27M subjects, a significant increase in the expression of FN1, COL1A1 and COL1A2 was also observed as compared to healthy controls (Fig. 7c–e). The increased gene expression for these ECM components in bulk RNA-seq from DMG H3K27M tumours further confirmed the observation made with FN1 and COL1A1 immunofluorescence analysis on archival DMG H3K27M tissue samples, that the DMG tumour microenvironment exhibits an increased expression for ECM components (Fig. 6; Suppl. Fig. 6).
Fig. 7: Validation of ECM component expression and prognostic value in DMG H3K27M human cohorts.The alternative text for this image may have been generated using AI.
a, b Gene expression levels for P2RY12 (a), and TMEM119 (b) in 16 DMG H3K27M subjects were extracted from the Gabriella Miller Kids First Pediatric Research Program (Kids First) cohort. The gene expression of these microglial markers in 60 spinal cord cervical C1 and 117 cerebellum brain region samples from healthy individuals, extracted from the Genotype-Tissue Expression Project (GTEx) cohort, were used for comparison. c–e Similar analyses as in panels a and b for FN1 (c), COL1A1 (d), and COL1A2 (e) gene expressions. f, g Gene expression levels for FN1 (f), COL1A1 (g), COL1A2 (h) and clinical data for a maximal 5-year period were extracted for 300 paediatric glioma subjects from the Children’s Brain Tumor Tissue Consortium (CBTTC) to assess their potential prognostic values. High and low expression were defined by the median value. Median expression for the gene of interest in 31 DMG, H3K27M subjects are indicated in the respective Kaplan-Meier 5-year survival curves. See Suppl. Fig. 7, for additional analyses performed on the Pacific Pediatric Neuro-Oncology Consortium (PNOC) and CBTTC human DMG cohorts.
To validate these findings across independent datasets, analyses were performed on two additional human cohorts from the Pacific Pediatric Neuro-Oncology Consortium (PNOC) and the Children’s Brain Tumor Tissue Consortium (CBTTC), open-access biorepositories with clinical and molecular data from paediatric brain tumours [47, 48]. Gene expression for FN1, COL1A1, and COL1A2 were found to be elevated in brain neoplasms defined as DMG/DIPG H3K27M in PNOC cohort and DMG H3K27M in the CBTTC cohort at diagnosis, reported as initial tumours (Suppl. Fig. 7b, c). The CBTTC cohort also included data for progressive tumours, which showed similar expressions for FN1, COL1A1, and COL1A2, as compared to the initial tumours, suggesting that the increase in expression of these ECM components observed in DMG tumours could be an early event in the physiopathology of these paediatric brain neoplasms (Suppl. Fig. 7c).
To assess the potential clinical relevance of these ECM components, we conducted survival probability analysis in the CBTTC cohort, which includes longitudinal data for 300 paediatric glioma patients, including 31 DMG H3K27M subjects. This analysis demonstrated that an increased FN1 expression was significantly associated with a worse 5-year survival (Fig. 7f), whereas increased COL1A1 and COL1A2 expressions exhibited a trend towards a negative correlation with prognosis, approaching statistical significance (Fig. 7g, h). In this human cohort, all DMG H3K27M cases exhibited high expression levels of FN1, COL1A1 and COL1A2, as determined by the cohort’s median expression values, as shown in the corresponding panels. These clinical findings further support the role of ECM alterations in DMG progression and highlight FN1 as a potential prognostic marker for DMG H3K27M tumours.
