Towards a 3D spheroid system for modelling leukodystrophies 117 4 To further explore variability, we examined differences in percentage of cells per cluster. Projecting LD spheroids onto the HD UMAP revealed that several cell types were underrepresented or absent in LD samples (Fig. 3G). A mixed-effects model confirmed disease-specific differences in six out of the 27 clusters (Fig. 3H-J, Supplemental Fig. 3). Specifically, cycling radial glia were severely reduced in all LDs (Fig. 3H), while outer radial glia was notably reduced in 4H and GLD (Fig. 3I). Although neural progenitors (NPs) were only significantly affected in GLD, they were also reduced in 4H and CD (Fig. 3J). Inhibitory progenitors showed reductions in GLD (Supplemental Fig. P-Q), while excitatory NP1 cells were more abundant in CD compared to controls (Supplemental Fig. 3J). Combined, this confirms cellular heterogeneity in LDs and indicates that even at the cellular composition level, disease-specific differences are detectable. The absence of clustering by institute and experimental batch supports the model's plausibility for cross-disease comparisons, as known technical factors were effectively controlled. Nevertheless, findings need to be interpreted with caution, as we did not identify clear disease-specific clustering patterns either. ◀ Figure 2: Single cell transcriptomics to identify cellular composition in myelinating spheroids. A) UMAP showing all cellular clusters in healthy donor (HD) spheroids. B) UMAP showing all clusters are represented at both timepoints. C) Bubble plot indicating a selection of positive gene markers used for cluster annotation. Size of bubble indicates the log10(adjusted significant difference) compared to all other clusters. Shade represents log fold change.
RkJQdWJsaXNoZXIy MTk4NDMw