Axonal injury is a targetable driver of glioblastoma progression

Abstract

Glioblastoma (GBM) is an aggressive and highly therapy-resistant brain tumour^1,2. Although advanced disease has been intensely investigated, the mechanisms that underpin the earlier, likely more tractable, stages of GBM development remain poorly understood. Here we identify axonal injury as a key driver of GBM progression, which we find is induced in white matter by early tumour cells preferentially expanding in this region. Mechanistically, axonal injury promotes gliomagenesis by triggering Wallerian degeneration, a targetable active programme of axonal death³, which we show increases neuroinflammation and tumour proliferation. Inactivation of SARM1, the key enzyme activated in response to injury that mediates Wallerian degeneration⁴, was sufficient to break this tumour-promoting feedforward loop, leading to the development of less advanced terminal tumours and prolonged survival in mice. Thus, targeting the tumour-induced injury microenvironment may supress progression from latent to advanced disease, thereby providing a potential strategy for GBM interception and control.

Fig. 1. Tumour development occurs preferentially in the WM.

a, Time-course analysis of npp tumour development. Tumour cells are tdTomato⁺ (tdTom, red); MBP (cyan) denotes WM; and nuclei were counterstained with DAPI (blue). The dashed boxes denote regions shown at higher magnification in the insets. Scale bars, 1 mm (main image) and 100 μm (inset). n = 4 (early), n = 4 (intermediate), n = 4 (late) and n = 3 (terminal) mice. b, Quantification of the percentage of tdTomato⁺ tumour cells located in WM (blue dots) or GM (grey dots) within the striatum of npp tumour-bearing mice shown in a. Statistical analysis was performed using two-way analysis of variance (ANOVA) with Tukey’s multiple-comparison correction (comparison of the percentage of tdTomato⁺ tumour cells in WM versus GM); comparison of the distribution of tdTomato⁺ cells between early, intermediate, late and terminal stage. Data are mean ± s.d. n = 4 (early), n = 4 (intermediate), n = 4 (late) and n = 3 (terminal) mice. P = 0.0041 (intermediate versus late), P < 0.0001 (all other comparisons). c, Quantification of the proportion of proliferating tdTomato⁺ cells in the WM or GM expressed as the percentage of Ki-67⁺tdTomato⁺ cells of total tdTomato⁺ cells in each region. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple-comparison correction. Data are mean ± s.d. n = 4 (early), n = 4 (intermediate), n = 4 (late) and n = 3 (terminal) mice. P < 0.0001 (early versus late and intermediate versus late), P = 0.003 (early versus terminal) and P = 0.0022 (intermediate versus late). ***P < 0.001, **P < 0.01. Source Data

Fig. 2. Developing tumours induce axonal injury in the WM.

a, Identification of myelinated regions. Top, H&E images of normal mouse brain (control) and the PDX model GCGR-L5 at the early (GCGR L5 early) and terminal (GCGR L5 terminal) stages. Bottom, as described for the top but overlayed with labels for high (green) and low (white) myelin marker genes expressing spots derived from ST data (Methods). Scale bars, 1 mm. b, Heat map of five k-mean clusters of log₂-transformed fold changes (log₂[FC]) from mouse genes significantly regulated between WM and GM in control, early and terminal PDX tumour ST spots (Methods). Selected terms enriched in cluster five are shown on the right. c, Gene set enrichment analysis in WM ST spots from control brain, or early and terminal PDX tumours. Normalized enrichment scores (NESs) of significantly enriched GO terms are shown (adjusted P < 0.1). d, Gene set enrichment analysis as in c, for cell type markers from ref. . The median of enrichment values across PDX models is shown. Astro, astrocytes; DC, dendritic cells; EC, endothelial cells; oligo, oligodendrocytes. e, The proportion of myelin high/low spots in bins of increasing tumour density (top). Middle, the proportion of spots assigned to anatomic brain regions in bins of increasing tumour density. Bottom, the average gene expression of functional categories in spots in bins of increasing density. f, The proportion of different cell types in bins of increasing tumour density derived by deconvolution using published scRNA-seq datasets (Methods). g, Reanalysis of ST data from ref. . Top, the proportion of spots assigned to histological regions in bins of increasing tumour density. Bottom, the averaged gene expression of different functional categories in bins of increasing tumour density.

Fig. 3. Axonal injury is an early event in gliomagenesis.

a,b, Fluorescence images of the contralateral (contra) or tumour-involved striatum (a) and quantification of the YFP mean fluorescence intensity (MFI) in striatal WM (b) of Thy1-YFP mice bearing early, intermediate, late and terminal npp tumours. The arrowheads indicate high-infiltrated (yellow, Thy1-YFP loss) and low-infiltrated (white, Thy1-YFP present) WM. For a, scale bars, 100 μm. Data are mean ± s.d. normalized to contralateral YFP MFI. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple-comparison correction. n = 4 (early, intermediate and late) and n = 3 (terminal) mice. c, Electron micrographs of tumour-involved and contralateral striatal WM of WT and Sarm1^−/− mice bearing intermediate npp tumours. The yellow arrows indicate degenerating axons. Scale bars, 2.5 μm. n = 4 (WT) and n = 4 (Sarm1^−/−) mice. d, Quantification of degenerating neurons in tumour-involved striatal WM in mice from c. Each dot represents a bundle. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple-comparison correction. Data are mean ± s.d. n = 4 (WT) and n = 4 (Sarm1^−/−) mice. P < 0.0001 for all comparisons. e, The g-ratios of tumour-involved (red dots) or contralateral (grey dots) striatal WM of mice from c. Each dot represents a WM bundle. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple-comparison test. Data are mean ± s.d. n = 4 (WT) and n = 4 (Sarm1^−/−) mice. f, Super-resolution images of npp tumour-bearing brains stained for mitochondrial marker TOMM20 (grey), tdTomato⁺ tumour cells (red) and axons (green). The yellow arrowheads indicate a mitochondria-laden varicosity, and the white arrowhead indicates a kinked axon. The side panels are orthogonal views indicating direct tumour cell–axonal contact (open and black arrowheads). Scale bars, 10 μm. n = 5 mice. g, The percentage of varicosities within 5 μm of a tumour cell body, cell process or >5 μm away from a tumour cell (no tumour cell). h, Atomic-force microscopy measurements of tissue stiffness (kPa) in tumour-involved (tumour) or contralateral WM of Thy1-YFP npp tumour mice (n = 5). Each spot represents force per indentation. Data are mean ± s.e.m. Statistical analysis was performed using two-sided Mann–Whitney U-tests. P = 0.0108. Source Data

Fig. 4. Transection of WM axons accelerates tumour progression.

a, Schematic of the experimental outline. w, weeks. b, Representative images of WT npp tumours subjected to sham treatment or injury at the intermediate tumour stage. tdTomato (red), EdU (grey) and DAPI (blue) are visualized. Scale bars, 1 mm. n = 7 mice per group. c, Quantification of the percentage of EdU⁺tdTomato⁺ tumour cells in WT and Sarm1^−/− npp tumours excluding the injury site (excl. inj. site) over the time course shown in a. Data are mean ± s.e.m. Statistical analysis was performed using multiple two-sided unpaired t-tests, with no adjustment for multiple comparisons. n = 6 (WT sham), n = 6 (WT injury), n = 6 (Sarm1^−/− sham) and n = 7 (Sarm1^−/− injury) at 4.5 weeks; n = 7 (WT sham), n = 7 (WT injury), n = 7 (Sarm1^−/− sham) and n = 6 (Sarm1^−/− injury) at 8.5 weeks; n = 7 (WT sham), n = 6 (WT injury), n = 7 (Sarm1^−/− sham) and n = 6 (Sarm1^−/− injury) mice at 12.5 weeks. d, Immunofluorescence images of tdTomato (red), GFAP (turquoise), CD68 (yellow) and IBA1 (magenta) staining in sham- or injury-group intermediate WT npp tumours. The dotted lines demarcate the corpus callosum (cc). str, striatum. Scale bars, 200 μm. n = 6 mice per group. e,f, Time-course analysis of the GFAP area (e) and CD68 intensity (IntDen; f) within WT and Sarm1^−/− npp tumours excluding the injury site. The fold change is relative to corresponding sham-treated tumour mice at each timepoint. Data are mean ± s.e.m. Statistical analysis was performed using multiple two-sided unpaired t-tests with no adjustment for multiple comparisons comparing injury to sham at each time point and per genotype. n = 6 (WT sham), n = 6 (WT injury), n = 6 (Sarm1^−/− sham) and n = 9 (Sarm1^−/− injury) at 4.5 weeks; n = 7 (WT sham), n = 6 and 8 (WT injury), n = 8 and 7 (Sarm1^−/− sham) and n = 8 (Sarm1^−/− injury) at 8.5 weeks; n = 7 (WT sham), n = 8 (WT injury), n = 5 (Sarm1^−/− sham) and n = 6 and 7 (Sarm1^−/− injury) mice at 12.5 weeks. g, Kaplan–Meier curves of npp tumour-bearing WT mice subjected to sham (WT sham) or injury (WT injury) at the intermediate disease stage. Statistical analysis was performed using log-rank tests. n = 15 mice for both groups. Median survival: 110 days (WT sham) and 92 days (WT injury) after electroporation. Source Data

Fig. 5. Sarm1 deletion inhibits GBM progression and ameliorates neurological function.

a–c, Representative images of tdTomato⁺ terminal WT and Sarm1^−/− npp tumours (a), quantification of the proportions of tumours with defined bulk (localized) or diffuse phenotype (diffuse) (b) and the tumour cell density in each genotype (c). For a, scale bars, 1 mm. Data are mean ± s.d. Statistical analysis was performed using two-sided unpaired t-tests. n = 10 (WT) and n = 9 (Sarm1^−/−) mice. d, Quantification of the tdTomato⁺ area in WT and Sarm1^−/− terminal tumours. Data are mean ± s.d. Statistical analysis was performed using two-sided unpaired t-tests. n = 10 (WT) and n = 6 (Sarm1^−/−) mice. e, Uniform manifold approximation and projection (UMAP) of scRNA-seq data from terminal WT and Sarm1^−/− npp tumours: neural progenitor-like (NPC-like), OPC-like, astrocyte-like (AC-like), MES-like and aNSC-like. f, As in e, but for microenvironmental cells: choroid plexus cells (CP), astrocytes, inflamed glia (infl. glia), OPCs, transient amplifying progenitors/neuroblasts (TAP/NB), aNSCs, ependymal cells (EpC), endothelial cells, pericytes, TAMs, monocytes (Mn) and T cells. g,h, The proportion of subpopulations between genotypes in tumour (g) and non-tumour (h) cell populations. The dashed line denotes equal proportions. Pearson’s χ² test < 0.05 and relative difference > 10% were considered to be significant (indicated by the hash symbol (#)). i–m, Flow cytometry analysis of immune populations (CD45 (i), TAMs (j), microglia (k), macrophages (l) and lymphocytes (m)) in terminal WT and Sarm1^−/− npp tumours. Data are mean ± s.d. Statistical analysis was performed using two-sided unpaired t-tests. n = 5 (WT) and n = 4 (Sarm1^−/−) mice. n, Kaplan–Meier analysis of npp tumour-bearing WT (grey) and Sarm1^−/− (turquoise) mice. Median survival: 125 (WT) and 148 (Sarm1^−/−) days. Statistical analysis was performed using log-rank tests. n = 22 (WT) and n = 18 (Sarm1^−/−) mice. o, Neuroscores of npp tumour-bearing WT (grey) and Sarm1^−/− (turquoise) mice at the indicated timepoints. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple-comparison correction. Data are mean ± s.d. n = 5 (WT) and n = 5 (Sarm1^−/−) mice. p, As for n, but for Sarm1-WT (WT, grey) and Sarm1^em1.1Tftc (turquoise) mice. Median survival: 120 (WT) and 152.5 (Sarm1^em1.1Tftc) days. Black lines denote censored animals. Statistical analysis was performed using log-rank tests. n = 10 (WT) and n = 12 (Sarm1^em1.1Tftc) mice. q, As described for o, but for Sarm1-WT (grey) and Sarm1^em1.1Tftc (turquoise) mice. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple-comparison correction. Data are mean ± s.d. Early: n = 8 (WT), n = 13 (Sarm1^em1.1Tftc); advanced: n = 6 (WT) and n = 11 (Sarm1^em1.1Tftc) mice. Source Data

Extended Data Fig. 1. Early tumour cells home to white matter.

a, Schematic of constructs used for in vivo electroporation in this study. From top to bottom: piggyBase. Basal tdTomato control piggyBac construct (tdTom) for integration of tdTomato alone; npp piggyBac construct for tumour induction. Adapted from (Clements et al.). b, Quantification of the percentage of striatal area occupied by MBP⁺ white matter in mice bearing npp tumours at indicated stages of tumour development. Tumour-involved striatum was quantified. One-way ANOVA with Tukey’s multiple comparisons. Mean ± SD, early, intermediate and late, n = 4 mice; for terminal, n = 3 mice. p = 0.015 (early vs late); p = 0.0333 (intermediate vs late). c, Time course analysis of SVZ/striatal brain regions of mice electroporated with tdTom construct. Representative images confirm efficient targeting and minimal migration in white matter regions adjacent to the SVZ in the absence of mutations. Scale bar=500 μm. Representative of n = 3 mice for each time point. d, Representative image of an olfactory bulb from a mouse collected at 4 weeks post-electroporation shown in c, confirming that tdTom-electroporated NSCs predominantly give rise to new olfactory neurons. Scale bar=500 μm. Representative of n = 3 mice. e, Representative images of tdTomato⁺ (red) npp tumours quantified in Fig. 1c. Tumours were collected at indicated stages and stained for the proliferation marker Ki67⁺ (grey), white matter marker (myelin basic protein, MBP, green) and DAPI (blue). Arrowheads highlight examples of tumour cell proliferation within (white) and outside (yellow) white matter. Scale bar=100 μm. Representative of early n = 4, Intermediate n = 4, Late n = 4, Terminal n = 3 mice. f, Representative images of indicated GFP-labelled PDX tumours collected early or at terminal disease (corresponding to 56 and 190 days for GBM2, and 70 and 143 for GL23, 34 and 56 days for GCGR E43 and 60 and 89 days for GCGR L5, respectively), stained for MBP (red). Dashed boxes denote regions shown at higher magnification in inset. Scale bar=500 μm. Images are representative of the following numbers of mice; GBM2 n = 3 early, n = 4 late; GL23 n = 3 early, n = 4 late; GCGRL5 n = 3 early, n = 3 late; GCGR E43 n = 3 early, n = 5 late. g, Quantification of percentage of GFP⁺ tumour cells located in white or grey matter within the striatum of PDX models shown in f. Comparison between %GFP⁺ cells in white matter and grey matter: two-way ANOVA with Tukey’s multiple comparisons. p < 0.0001 (early WM vs early GM). Comparison of the distribution of GFP⁺ cells (calculated as %GFP⁺ cells in WM - %GFP⁺ cells in GM) between early and terminal stages: Two-sided paired t test.; p = 0.0127 (early vs terminal). Mean ± SD. Early n = 4, Terminal n = 4 mice. Source Data

Extended Data Fig. 2. Spatial transcriptomics analysis of PDX models of glioblastoma.

a, H&E staining of spatial transcriptomics slides with two sections each (S1 and S2) from six early tumours, eight advanced PDX tumours and one control tumour-free brain (Control). b, Schematic diagram of alignment and filtering strategy of sequencing reads. Reads were mapped to three genomes: mouse, human and a combined genome of both species. Only reads mapping confidently to a single species were kept for further analysis. c, Distribution of the ratio of human to mouse reads in section 1 (S1) and section 2 (S2) across the dataset. Number of spots was n = 3412 for S1 and n = 3301 for S2. The centre line in the box plot represents the median, the lower and upper hinges correspond to the first and third quartiles. The upper whisker extends from the hinge to the largest value no further than 1.5 x IQR (interquartile range) from the hinge. The lower whisker extends from the hinge to the smallest value at most 1.5 x IQR of the hinge. Outliers are plotted as individual dots. d, Distribution of the ratio of human to mouse reads in control mice. e, Spatial distribution of tumour density ratios. Top two rows: Spatial plots of tumour densities from section 2 in the 8 terminal PDX tumours; Bottom row: Example of H&E image-based quantification of tumour density in normal and tumour regions (see Methods). Left panels are H&E images and right panels are processed images with nuclei masks in yellow and ST spots position circled in turquoise. f, Mean tumour density inferred from nuclei density as a function of tumour density measures inferred from sequencing data (tumour density ratios, see Methods). g, Systematic analysis of expression trends with tumour density. Heatmap of S statistics from Mann-Kendall tests for expression of GO categories as a function of tumour density in the entire dataset. Values were clustered using K-means. h-o, Related to Fig. 2e, expression trends of gene signature as function of tumour density for terminal PDX tumours. Black lines represent trends in control pseudo-spots (see Methods). Error bars were calculated as mean ± SD across pseudo spots. p, Spatial plots of ST data in control brain and the 8 terminal PDX tumours (first row is section 1 (S1) and second row section 2 (S2)), as indicated. Seven anatomical regions were annotated and are displayed in different colours (see Methods). q, Re-analysis of the Moffet et al CosMx dataset. Heatmap of gene signatures expression trends as a function of tumour density in patient GBM. Expression values were z-score normalized across tumour density bins.

Extended Data Fig. 3. Axonal injury increases with tumour cell density.

a, YFP mean fluorescence intensity (MFI) in the tumour-involved striatum normalized to intensity in contralateral striatum and plotted as a function of number of tumour cells in the white matter. Each spot represents a mouse at early (light green), intermediate (turquoise), late (dark blue) and terminal (black) stages. Two-tailed Pearson R correlation. p < 0.0001. For early, intermediate and late n = 4, for terminal n = 3 mice. b, Representative images of white matter bundles in the tumour-involved (Tumour) and contralateral (tumour-free; Contra) striatum of Thy1YFP mice bearing intermediate npp tumours. Arrowheads exemplify a heavily- (yellow), moderately- (white) and a lowly- (blue) bundle infiltrated by tdTomato⁺ tumour cells (red). Scale bar=100 μm. c, Quantification of YFP mean fluorescence intensity (MFI) as a function of tumour cell density in tumour-involved striatal white matter of npp tumour-bearing mice from b. Turquoise line and blue band indicate mean contralateral fluorescence intensity and SD, respectively. Two-tailed Pearson R correlation. n = 6 mice. Each spot represents an individual bundle. d, Representative images of neurofilament staining (NF, yellow) of tumour-involved (tumour) or contralateral (tumour-free, Contra) striatal white matter bundles in WT and Sarm1^-/- mice bearing intermediate npp tumours. Moderately- (white arrowhead) and heavily- (yellow arrowhead) infiltrated white matter bundles are highlighted. Scale bar=50 μm. e, Quantification of neurofilament (NF) mean fluorescence intensity (MFI) in tumours depicted in d. Individual white matter bundles are shown. Two-way ANOVA with Tukey’s multiple comparisons. Mean ± SD, WT n = 3, Sarm1^-/- n = 3 mice. p < 0.0001 (WT contralateral vs WT tumour); p < 0.0001 (WT tumour vs Sarm1^-/- tumour). f, Representative images correlating confocal microscopy images and electron micrographs of intermediate tumours (10.5 weeks post-electroporation) induced in WT (i-iv) and Sarm1^-/- mice (v-viii); i, overview image of npp tumours in WT and v, Sarm1^-/- mice (bottom). Scale bar=200 μm. Representative of WT n = 4, Sarm1^-/- n = 4 mice. ii and vi, Corresponding representative electron micrographs of fluorescence images from i and v. Scale bar=200 μm. Dashed boxes indicate the striatal white matter bundle depicted at higher magnification on the right (iii, iv, vii and viii). Dashed yellow boxes indicates region shown in Fig. 3c. Scale bar=50 μm. g, G-ratios of pathological or intact axons in the tumour-involved striatal white matter of WT mice bearing intermediate npp tumours. Each dot represents an axon. Mean ± SD. Two-sided unpaired t test. n = 4 mice. h-j, Representative immunofluorescence images of intermediate npp tumours generated in Thy1YFP mice and stained for indicated markers of proteinopathies. Bottom panels in h and I indicate successful antibody staining in positive control tissue. Scale bar=200 μm for h and I, Scale bar=50 μm for j. k, Representative immunofluorescence images of intermediate npp tumours generated in WT mice (n = 3 mice) and stained for Hypoxyprobe following pimonizadole administration. Terminal tumours (bottom panels) served as positive control (n = 3 mice). Scale bar=200 μm.l, Representative images of tdTomato⁺ (red) intermediate npp Thy1YFP tumours stained for phospho-myosin light chain 2 (pMLC2) (grey). Shown are tumour-involved (Tumour) and contralateral (Contra) white matter. Scale bar=100 μm. Images are representative of n = 3 mice. Source Data

Extended Data Fig. 4. Axonal injury is accompanied by neuroinflammation.

a-c, Representative images of GFAP (magenta) and CD68 (turquoise) staining in brains from age-matched healthy control mice or within early, intermediate, late and terminal npp tumours generated in WT mice. Scale bar=500 μm. Control n = 6, early n = 6, intermediate n = 7, late n = 7 and terminal n = 4 mice. b-c, Quantification of GFAP area (b) and CD68 integrated density (c) within the tumour region of samples from a, plotted as fold change over levels in brain region-matched controls. Mean ± SEM. Multiple two-sided unpaired t tests. Control n = 6, early n = 6, intermediate n = 7, late n = 7 and terminal n = 4 mice. In b, p < 0.0001 for control vs early, control vs intermediate and control vs terminal; p = 0.0004 control vs late; p = 0.0281 early vs late; p = 0.0217 early vs terminal; p = 0.0494 intermediate vs late. In c, p < 0.0001 for control vs terminal and intermediate vs terminal; p = 0.0024 control vs early; p = 0.0150 control vs intermediate; p = 0.0023 control vs late; p = 0.0005 early vs terminal; p = 0.0112 intermediate vs late. d, Representative images of GFAP⁺ astrocytes surrounding tumour-involved striatal white matter bundles in mice bearing intermediate npp tumours. Contralateral tumour-free striatum (Contra) is shown on the right. Scale bar=100 μm. Images are representative of n = 6 mice. e, Quantification of GFAP⁺ astrocyte density as a function of tumour cell density in tumour-involved striatal bundles in mice bearing intermediate npp tumours. Each point represents a bundle, ≥5 bundles in tumour-involved striatum per mouse were quantified. Turquoise line and blue band indicate mean contralateral GFAP⁺ cell density and SD, respectively. Two-tailed Pearson R correlation. p < 0.0001. n = 6 mice. f, Representative images of CD68⁺/Iba1⁺ microglia in tumour-involved striatal white matter bundles in mice bearing intermediate npp tumours. Contralateral tumour-free striatum (Contra) is shown on the right. Scale bar=100 μm. Images are representative of n = 4 mice. g, Quantification of CD68 integrated intensity (IntDen) as a function of tumour cell density in tumour-involved striatal bundles in mice bearing intermediate npp tumours. Each point represents a bundle, ≥5 bundles in tumour-involved striatum per mouse were quantified. Turquoise line and blue band indicate mean contralateral CD68 integrated density and SD, respectively. ±1 SD range. Two-tailed Pearson R correlation. p < 0.0001. n = 4 mice. h-l, Flow cytometry analysis of indicated immune populations in age-matched control healthy brains or early, intermediate and late npp tumour generated in WT mice. Mean ± SD. For control, early, and late n = 4, for intermediate n = 3 mice. Multiple two-sided unpaired t tests. In h; p = 0.0011 (control vs early); p = 0.0062 (control vs intermediate); p < 0.0001 (control vs late); p = 0.0062 (early vs intermediate); p < 0.0001 (early vs late). In i; p = 0.0037 (control vs intermediate); p < 0.0001 (control vs late); p = 0.0041 (early vs intermediate); p < 0.0001 (early vs late). In j; p = 0.0124 (control vs intermediate); p = 0.001 (control vs late); p = 0.0151 (early vs intermediate); p = 0.0012 (early vs late). In k; p < 0.0001 (control vs intermediate); p = 0.0008 (control vs late); p < 0.0001 (early vs intermediate); p = 0.0008 (early vs late); p = 0.0097 (late vs terminal); In l; p = 0.0353 (control vs intermediate); p = 0.0078 (control vs late); p = 0.0434 (early vs intermediate); p = 0.0092 (early vs late). m-n, Quantification of GFAP area (m) CD68 integrated density (n) within the tumour region of indicated early and terminal PDX models plotted as fold change relative to contralateral tumour-free brain tissue. Each point represents a tumour. n = 5 mice. Mean ± SEM. Multiple two-sided paired t tests. In m, p = 0.0023 for contralateral vs early; p = 0.0016 for contralateral vs terminal. Source Data

Extended Data Fig. 5. Inactivation of Sarm1 preserves axonal integrity and suppresses inflammation in npp tumours following experimental injury.

a, Representative images of intermediate npp tumours generated in WT and Sarm1^-/- mice, stained for neurofilament (NF, yellow) 2 weeks following Sham injury or transection of tumour-ipsilateral corpus callosum axons (Injury, see timeline in Fig. 4a). Scale bar=1 mm. White box indicates the injury site region depicted at high magnification on the right. Note significant axonal protection in Sarm1^-/- animals even at the site of injury demarcated by the star symbol. Scale bar=200 μm. Images are representative of n = 3 mice per condition. b, Representative immunofluorescence images of npp tumours generated in WT and Sarm1^-/- animals and subjected to Sham or Injury at 4.5 weeks, 8.5 weeks or 12.5 weeks after tumour induction (see timeline in Fig. 4a). tdTomato⁺ tumour cells are in red; sections were stained for EdU (grey) and DAPI (blue). Scale bar=200 μm. WT Sham n = 6, WT Injury n = 6, Sarm1^-/- Sham n = 6, Sarm1^-/- Injury n = 7 at 4.5 weeks, WT Sham n = 7, WT Injury n = 7, Sarm1^-/- Sham n = 7, Sarm1^-/- Injury n = 6 at 8.5 weeks; WT Sham n = 7, WT Injury n = 6, Sarm1^-/- Sham n = 7, Sarm1^-/- Injury n = 6 mice at 12.5 weeks. c, Quantification of the percentage of EdU⁺ tumour cells at the injury site in samples from b. Shown is the fold change relative to corresponding brain region in Sham tumour mice at each time point. Mean ± SEM. Multiple two-sided unpaired t tests. WT Sham n = 6, WT Injury n = 6, Sarm1^-/- Sham n = 6, Sarm1^-/- Injury n = 7 at 4.5 weeks, WT Sham n = 7, WT Injury n = 7, Sarm1^-/- Sham n = 7, Sarm1^-/- Injury n = 6 at 8.5 weeks; WT Sham n = 7, WT Injury n = 6, Sarm1^-/- Sham n = 7, Sarm1^-/- Injury n = 6 mice at 12.5 weeks. p = 0.0094 (Sarm1^-/- sham vs Sarm1^-/- injury site). d, Representative images of GFAP (turquoise), Iba1 (magenta) and CD68 (yellow) staining within the injury site of npp-tumour bearing WT mice subjected to corpus callosum transection injury at 8.5 weeks post tumour induction. Scale bar=200 μm. Sham n = 6, WT Injury n = 6, Sarm1^-/- Sham n = 6, Sarm1^-/- Injury n = 9 at 4.5 weeks, WT Sham n = 7, WT Injury n = 6 (e) and 8 (f), Sarm1^-/- Sham n = 8 (GFAP) and 7 (Iba1/CD68), Sarm1^-/- Injury n = 8 at 8.5 weeks; WT Sham n = 7, WT Injury n = 8, Sarm1^-/- Sham n = 5, Sarm1^-/- Injury n = 6 (GFAP) and 7 (Iba1/CD68) mice at 12.5 weeks. e-f, Quantification of GFAP area (e) and CD68 intensity (f) at the injury site of samples described in b. Shown is the fold change relative to corresponding brain region in Sham tumour mice at each time point. Mean ± SEM. Multiple two-sided unpaired t tests. In e, p < 0.0001 at both 4.5 and 8.5wk WT Sham vs WT Injury (injury site). In f, p = 0.012 (4.5wk) and p = 0.0016 (8.5wk) WT Sham vs WT Injury (injury site). Sham n = 6, WT Injury n = 6, Sarm1^-/- Sham n = 6, Sarm1^-/- Injury n = 9 at 4.5 weeks, WT Sham n = 7, WT Injury n = 6 (e) and 8 (f), Sarm1^-/- Sham n = 8 (e) and 7 (f), Sarm1^-/- Injury n = 8 at 8.5 weeks; WT Sham n = 7, WT Injury n = 8, Sarm1^-/- Sham n = 5, Sarm1^-/- Injury n = 6 (e) and 7 (f) mice at 12.5 weeks. Source Data

Extended Data Fig. 6. Neuron-specific Sarm1 inactivation delays tumour development.

a, Schematic showing intraventricular injection of AAV8-Syn-EGFP (GFP) or AAV8-Syn-SARM1-CDN-EGFP (SarmDN) alongside npp tumour inducing plasmids at P2. b, Representative images of the injury site in npp tumour-bearing brains transduced with GFP or SarmDN AAVs at P2 and subjected to axonal transection at intermediate disease stage. Tissue was stained for neurofilament (grey). GFP (green) denotes successful neuronal transduction. Note axonal protection in the SarmDN-transduced brains. Scale bar=200 μm. GFP n = 3, SarmDN n = 3 mice. c, Representative images of tdTomato (red) and GFP (green) fluorescence in npp tumours injected intraventricularly with GFP or SarmDN AAVs at the time of tumour induction (P2), subjected to Sham (left panel) or Injury (right panel) at intermediate stage and stained for EdU (grey) and DAPI (blue). Scale bar=200 μm. GFP Sham n = 5, GFP Injury n = 5, SarmDN Sham n = 4, SarmDN Injury n = 5 mice. d, Quantification of the percentage of proliferating EdU⁺/tdTomato⁺ tumour cells over total number of tdTomato⁺ tumour cells in tumours from c. Mean ± SEM. Multiple two-sided unpaired t tests. GFP Sham n = 5, GFP Injury n = 5, SarmDN Sham n = 4, SarmDN Injury n = 5 mice. p = 0.0411 (GFP injected, sham vs injury (excluding injury site); p = 0.0309 (GFP injected, Injury (excluding injury site) vs Injury (injury site), p = 0.0043 (GFP Injury (excluding injury site) vs SarmDN Injury (excluding injury site). e-f, Quantification of GFAP area (e) and CD68 intensity (IntDen, f) within the tdTomato region in tumours from c; injury site and the rest of the tumour area were quantified separately and normalized to Sham in each genotype. Mean ± SEM. Multiple two-sided unpaired t tests. GFP Sham n = 7, GFP Injury n = 5, SarmDN Sham n = 4, SarmDN Injury n = 5 mice. In e, p = 0.02 (GFP injected, Sham vs Injury (excluding injury site); p = 0.0012 (GFP injected, Sham vs Injury (injury site). g, Schematic showing intratumoural injection of AAV8-Syn-EGFP (GFP) or AAV8-Syn-SARM1-CDN-EGFP (SarmDN) at intermediate/late time point. h, Representative images of tdTomato (red) and GFP (green) fluorescence in WT npp tumours injected intratumorally with GFP or SarmDN AAVs at intermediate disease stage and stained for EdU (grey). Scale bar=200 μm. Images are representative of 4 mice per condition. i, Quantification of percentage EdU⁺/tdTomato⁺ tumour cells over total number of tdTomato⁺ tumour cells in tumours from h. Mean ± SEM. Two-sided unpaired t test. GFP n = 4, SarmDN n = 4 mice. p = 0.0248 j, as in h for tumours injected with AAVs at late stage. Scale bar=200 μm. Images are representative of 5 GFP mice and 4 SarmDN mice. k, Quantification of percentage EdU⁺/tdTomato⁺ tumour cells over total number of tdTomato⁺ tumour cells in tumours from j. Mean ± SEM. Two-sided unpaired t test. GFP n = 5, SarmDN n = 4 mice. Source Data

Extended Data Fig. 7. Wallerian degeneration contributes to tumour cell tropism to the white matter.

a, Quantification of the number of tdTomato⁺ tumour cells located in white matter (WM) or grey matter (GM) within the striatum of Sarm1^-/- intermediate npp tumours. Mean ± SD. Two- sided unpaired t test. p < 0.0001, n = 5 mice. b-f, Summary statistics generated from multiple simulations of the agent-based model. To set the timescales for the simulations, tumour progression for a total of 100 survival days in the WT and 150 survival days for the Sarm1^-/- was assumed. b, Number of tumour cells occupying the white and grey matter over time. c, Relative proportions of tumour cells occupying the white and grey matter over time. d, Proliferation rates of tumour cells occupying the white and grey matter over time. e, Normalized total levels of axonal injury over time. f, Spatial tumour density profiles at terminal stage of WT and Sarm1 mutant, computed as a moving average of tumour cell numbers located in 5 × 5 square regions. g, Example snapshots of agent-based model simulations. Snapshots of glioma progression at early, intermediate, late and terminal timepoints are shown for WT and Sarm1^-/- tumours. Grey matter is shown in grey colour; white matter bundles are shown as white ellipses and tumour cell occupancy is shown in red scatter points. Source Data

Extended Data Fig. 8. Neuropathological assessment of npp tumours generated in wild-type and Sarm1^-/- mice.

a, Representative H&E wholemount images of terminal tumours generated in WT (i-v) and Sarm1^-/- mice (vi-ix). Scale bar=2.5 mm (i, vi); =500 μm (iv, v and vii); = 250 μm (ii, iii); 100 μm (vii) 50 μm (ix). Black arrowheads indicate mitotic activity. Images are representative of 15 WT and 8 Sarm1^-/- mice. b, Representative immunofluorescence images of npp terminal tumours generated in wild-type (WT) and Sarm1^-/- mice and stained for Ki67 (grey). Tumour cells are identified by tdTomato fluorescence and nuclei are counterstained with DAPI (blue). Scale bar=200 μm. c, Quantification of percentage of Ki67⁺ tumour cells in tumours shown in b. n = 4 animals for both genotypes. Mean ± SEM. Two-sided unpaired t test. d, Representative immunofluorescence images of npp terminal tumours generated in WT and Sarm1^-/- mice and stained for the endothelial marker CD31 (grey). Tumour cells are identified by tdTomato fluorescence (red) and nuclei are counterstained with DAPI (blue). Scale bar=200 μm. e-h, Quantification of indicated vascular phenotypes in tumours from d; n = 4 animals for each genotype. Mean ± SEM. Two-sided unpaired t test. in g, p = 0.0098; in h, p = 0.0464 i, Representative immunofluorescence images of npp terminal tumours generated in WT and Sarm1^-/- mice and stained for the basal lamina marker laminin (red) and CD31 (green). Scale bar=200 μm. j, Quantification of the percentage of CD31⁺ vessels covered with laminin within tumours from i. n = 5 animals for each genotype. Mean ± SEM. Two-sided unpaired t test. k, Representative immunofluorescence images of npp terminal tumours generated in WT and Sarm1^-/- mice and stained for of the pericyte marker PDGFRB (red) and CD31 (green). Scale bar=200 μm. l, Quantification of the percentage of CD31⁺ vessels covered with pericytes within tumours from k. n = 4 animals for each genotype. Mean ± SEM. Two-sided unpaired t test. m, Representative images of IgG extravasation (green) in npp terminal tumours generated in WT and Sarm1^-/- mice. Tumour cells were identified by tdTomato fluorescence (red) and nuclei are counterstained with DAPI. Scale bar=200 μm. n, Quantification of IgG⁺ tumour area over total tumour area within tumours from m. WT n = 5, Sarm1^-/- n = 4 mice. Mean ± SEM. Two-sided unpaired t test. o, Representative immunofluorescence images of npp terminal tumours generated in WT and Sarm1^-/- mice and stained for Iba1 (green). Tumour cells are identified by tdTomato and nuclei are counterstained with DAPI (blue). Scale bar=200 μm. p, Quantification of the number of Iba1⁺ cells per mm² tumour area in tumours shown in o. n = 4 animals for both genotypes. Mean ± SEM. Two-sided unpaired t test. Source Data

Extended Data Fig. 9. Heterotypic signalling is reduced in Sarm1^-/- npp tumours.

a, UMAP representation of scRNA-seq data from microenvironmental cells identified as TAMs in Fig. 5f. Two clusters are labelled in red (1) and blue (2). b, Heatmap of logexpression ratios (logFC) between cluster 1 and 2 for microglia markers. c, as in b for macrophage markers. d, Proportion of microglia (MG) or macrophages (MAC) in tumours from WT or Sarm1^-/- mice. The dashed line represents equal proportions in both genotypes. Cell types with P_Pearson’s chi-squared test <0.05 and relative difference > 10% were considered significantly different (§). e, LIANA Ligand-receptor analysis based on ligands from microenvironmental cells (Sender) and receptors from tumour cells (Receiver) in tumours from WT animals. Numbers on the plot denote the numbers of significant interactions. f, As in e for Sarm1^-/- animals. g, LIANA Ligand-receptor analysis based on ligands from microenvironmental cells (Sender, cell types at the top of the graph) and receptors from tumour cells (Receiver, cell types at the bottom of the graph) in tumours from WT animals. The diameter of the dots represents the proportion of the cells expressing ligands among the sender cells (ligand.prop) and the colour magnitude is the mean of the ligand and the receptor gene expression (lr.mean). Ligand-receptor pairs are listed on the left of the graph. h, As in g for Sarm1^-/- animals. Source Data

Extended Data Fig. 10. Genetic inactivation of Sarm1 results in more diffuse tumours in a second independent mouse model.

a, Tumour cell density in terminal npp tumours generated in wild-type (grey, WT) or Sarm1^-/- (turquoise) mice plotted against survival. Circles indicate localized tumours; triangles indicate diffuse tumours. No correlation is found between tumour density and survival in either genotype. Two-sided Pearson correlation R. Regression line with 95% confidence interval per genotype. n = 10 for WT, n = 9 mice for Sarm1^-/-. b, Survival curves of WT tumours stratified on localized or diffuse histology. Mean ± SD survival for WT mice with a localized or diffuse tumour were 112 ± 12 and 112 ± 14, respectively. n = 11 for localized tumours and n = 9 mice for diffuse tumours. c, Representative fluorescence images of npp tumours generated in Sarm1^em1.1Tftc and genetic background matched Sarm1^wt mice (WT). Images show tdTomato⁺ tumour cells (red) and nuclei are counterstained with DAPI (blue). Scale bar=1 mm. d, Quantification of number of tumour cells per mm² in tumours shown in c. Mean ± SD. Two-sided unpaired t test. n = 4 mice for both genotypes. p = 0.042. Source Data