Spatial transcriptomic analysis of neurons at the glioma infiltration front: A case study Kaishi Satomi, Hirokazu Takami, Hiromi Onizuka, Kuniaki Saito, Takako Yoshioka, Motoo Nagane, Yutaka Suzuki, Koichi Ichimura, Junji Shibahara Abstract Spatial transcriptomics reveals localized IGFBP5 expression and TGFβ‐related signaling in neurons within glioma‐infiltrated tissue. Recent studies have demonstrated that functional neuron–glioma interactions contribute to glioma progression through synapse‐like communication [, , ]. However, previous studies have primarily examined the effects of neurons on tumor cells, whereas the effects of tumor cells on surrounding neurons remain incompletely understood. Spatial transcriptomics enables simultaneous evaluation of morphology and gene expression within preserved tissue architecture in human tumors []. The Xenium In Situ platform (10x Genomics, Pleasanton, CA, USA) enables visualization of transcript localization at near‐cellular resolution while preserving histopathological context, although important platform‐dependent analytical limitations remain []. Formalin‐fixed paraffin‐embedded tissue from a surgically resected tumor in a Japanese man in his 80s at Kyorin University Hospital was analyzed using the Xenium In Situ platform (10x Genomics) with the Human Brain Gene Expression Panel comprising 366 genes. The tumor was diagnosed as glioblastoma, IDH‐wildtype, based on the integrated histopathological and molecular findings according to the WHO Classification of Central Nervous System Tumors, 5th edition. DNA methylation profiling using the Epignostix CNS Tumor Classifier version 12.8.1 (Epignostix GmbH, Heidelberg, Germany) yielded a calibrated score of 0.88478 for “pediatric‐type diffuse high‐grade gliomas” but did not result in assignment to a specific methylation class. Therefore, we regarded this methylation result as supportive molecular information rather than sufficient evidence to reclassify the tumor as a specific pediatric‐type diffuse high‐grade glioma (Figure ). Copy number analysis further demonstrated PDGFRA amplification and homozygous deletion of CDKN2A and CDKN2B (Figure ). Data were analyzed using Xenium Explorer version 4.1.1 (10x Genomics), R version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria), utilizing Seurat version 5.1.0, and pathway enrichment workflows. To further evaluate the pSMAD2 findings, four additional adult high‐grade diffuse glioma cases with glioma‐infiltrated and remote cortical regions were examined using immunohistochemistry. pSMAD2‐positive neurons were quantified in 10 fields per region and compared using a two‐sided Mann–Whitney U test. The study was approved by the institutional ethics committee (R05‐188). Unsupervised clustering identified 18 clusters representing tumor cells, neurons, astrocytes, oligodendrocytes, endothelial cells, and immune‐associated cell populations (Figure ). The inferred cellular distributions were morphologically consistent with the infiltrative histological architecture of the tissue sections (Figure ). Tumor cell clusters demonstrated high PDGFRA expression (Figure ), consistent with PDGFRA amplification. Spatial transcriptomic visualization demonstrated distinct spatial localization patterns between Clusters 0 and 1 tumor cell populations, indicating spatial intratumoral transcriptional heterogeneity among PDGFRA‐expressing tumor cells (Figure ). TGFB2 was expressed in several tumor‐associated clusters, whereas POSTN was enriched in a distinct tumor cell subset corresponding predominantly to Cluster 1 (Figure ). Spatial transcriptomic visualization further demonstrated that POSTN‐expressing Cluster 1 tumor cells showed a distinct spatial distribution, with preferential localization in relatively deep tumor regions, including areas containing vascular structures (Figures and ). TGFB2‐associated TGFβ/BMP signaling pathways were enriched predominantly in Cluster 0 (Tables and ), although TGFB2 transcripts were detected across multiple tumor‐associated clusters (Figure ). We therefore examined whether neurons spatially associated with TGFB2‐enriched Cluster 0 tumor cells showed localized neuronal transcriptional alterations and TGFβ‐related signaling. Representative infiltrative and remote cortical regions are shown in Figure . Histological evaluation identified morphologically viable residual neurons entrapped within infiltrative tumor tissue (Figure ), whereas neurons in the remote cortex showed preserved architecture (Figure ). During exploratory spatial transcriptomic analysis, IGFBP5 was identified as preferentially expressed in tumor‐associated neuronal populations and was selected for further evaluation. IGFBP5 expression was mainly observed in SLC17A7‐positive neuronal populations within glioma‐infiltrated regions (Figure ). Among the neuronal markers examined, SLC17A7 showed the clearest correspondence with morphologically identifiable entrapped neurons. Combined visualization showed close spatial juxtaposition between SLC17A7‐positive/IGFBP5‐positive neurons and adjacent PDGFRA‐positive tumor cells (Figure ). In contrast, neurons in remote non‐infiltrated regions showed substantially lower IGFBP5 expression (Figure ). Quantitative comparison confirmed significantly higher IGFBP5 expression among SLC17A7‐positive neuronal populations in the analyzed infiltrated region than in the selected remote cortical region (Figure ). PDGFRA‐positive tumor cells infiltrated around these residual neurons, resulting in close spatial juxtaposition between tumor cells and IGFBP5‐positive neuronal populations. Notably, these neuronal populations were preferentially localized within tumor regions corresponding to Cluster 0, which demonstrated enrichment of TGFB2‐associated TGFβ/BMP signaling pathways. Proliferative activity appeared comparable among adjacent tumor cells. We next examined whether neurons within glioma‐infiltrated regions showed localized TGFβ‐related signaling. Immunohistochemical analysis identified pSMAD2‐positive neuronal populations adjacent to infiltrating tumor cells (Figure ), whereas neurons in remote non‐infiltrated regions lacked detectable pSMAD2 expression (Figure ). In serial sections, pSMAD2‐positive neurons and IGFBP5‐expressing SLC17A7‐positive neurons were observed in closely corresponding fields within the same glioma‐infiltrated regions. Because immunohistochemistry and Xenium analysis were performed on different sections, direct single‐cell‐level colocalization could not be confirmed. Nevertheless, their close spatial correspondence supports the interpretation that IGFBP5 expression and TGFβ‐related signaling occurred within the same tumor‐associated neuronal population. These findings suggest localized neuronal transcriptional state alterations associated with the glioma microenvironment. Four additional high‐grade diffuse glioma cases were examined for pSMAD2 expression. Two showed significantly higher proportions of pSMAD2‐positive neurons in the glioma‐infiltrated regions, whereas the other two did not (Figure ), indicating that localized neuronal SMAD2 phosphorylation occurred in a subset of cases. The present findings suggest that TGFβ‐related signaling may be active in tumor‐associated neurons within glioma‐infiltrated regions, as supported by pSMAD2‐positive neuronal populations adjacent to infiltrating tumor cells. Notably, these neurons represented morphologically viable residual neurons embedded within infiltrative tumor tissue, rather than degenerating neurons associated with necrotic regions. Thus, localized neuronal transcriptional alterations may occur in viable neurons exposed to the glioma microenvironment. IGFBP5 has been implicated in extracellular matrix remodeling and fibrosis‐associated transcriptional responses in multiple tissues and tumors []. Similarly, TGFβ signaling is a major regulator of extracellular matrix‐associated and mesenchymal‐like transcriptional programs and has been associated with mesenchymal states in gliomas []. Although interactions between TGFβ‐related pathways and IGFBP5‐associated biological processes have been suggested in fibrosis‐associated conditions [], the spatial association between TGFβ‐related signaling and neuronal IGFBP5 expression within infiltrative glioma tissue has not, to our knowledge, been previously described. The biological significance of these observations remains unclear and requires further investigation. This study had several limitations. First, Xenium‐based spatial transcriptomics demonstrates spatially associated transcriptional states within preserved tissue architecture but does not establish definitive molecular interactions or causal biological relationships. Second, the predefined and relatively restricted gene panel limited comprehensive ligand–receptor‐based cell–cell interaction analysis. Additionally, differences in spatial resolution among platforms may limit direct visualization of the cell‐level relationships observed here. Therefore, the observed spatial association among TGFB2‐expressing tumor cells, pSMAD2‐positive neurons, and IGFBP5‐expressing neurons should be interpreted as hypothesis‐generating rather than as definitive evidence of a specific signaling pathway. Third, the restricted gene panel did not permit comprehensive characterization of neuronal subtypes across brain regions. SLC17A7 was selected because it showed the clearest correspondence with morphologically identifiable residual neurons. Complementary immunohistochemical analysis of phosphorylated SMAD2 provided orthogonal evidence supporting localized TGFβ‐related signaling activity in tumor‐associated neurons. Collectively, neurons within glioma‐infiltrated tissue showed localized transcriptional alterations accompanied by evidence of TGFβ‐related signaling, highlighting the utility of Xenium‐based spatial transcriptomics for visualizing tumor‐associated neuronal states within preserved infiltrative glioma architecture. Integrating spatial transcriptomics with morphology‐ and protein‐based pathological analyses may provide a valuable hypothesis‐generating framework for investigating localized tumor microenvironmental interactions in human gliomas. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest.