Fig. 4: Neuronal cholesterol receptors mediate pro-synaptic effects of astrocytes.

a, Experimental overview of the CRISPRi experiment. We used three different gRNAs to knock down two cholesterol receptor genes in neurons, LRP1 and LDLR, along with three non-targeting (NT) guides to serve as control samples. The knockdown NPCs were plated with mouse astrocytes and cultured for 48 hours, after which they were harvested for RNA-seq. b, Boxplots of normalized expression (logCPM) of LRP1 (left) and LDLR (right) show effective knockdown (KD) of the target genes by CRISPRi. The p-values were calculated using DREAM linear mixed models. c-d, Scatter plots of logCPM values from the differential expression (DE) analysis of the KD samples. The diagonal (y = x) is indicated with a black line. Correlation coefficients (R) and p-values were computed using Pearson correlation where LRP1 KD is teal and LDLR KD is pink. The DE analysis yielded similar results; the fold changes of the 1,774 shared significantly DE genes (c), as well as 1,858 unique significantly DE genes (d) were both highly correlated (R = 0.93 and p < 2.2e-16, R = 0.78 and p < 2.2e-16, respectively). e, A venn diagram of the neuronal synaptic genes downregulated in LRP1- and LDLR-KD neurons (n = 245, shown in beige) and neuronal synaptic genes upregulated with astrocyte LF2 (n = 533, shown in green) have an intersection of 62 genes (Fisher’s exact test OR = 4.637, p = 1.653e-19). f, Heatmap of standardized expression (z-score logCPM) of the 62 synaptic genes show a similar downregulation in both LRP1-KD and LDLR-KD against NT-KD samples. g, Boxplots of normalized expression (logCPM) of two SNAP-n genes, DRD2 and GAP43, in LRP1-KD, LDLR-KD, and NT-KD neurons. The p-values were calculated using linear mixed models in the DREAM software package. The boxplots’ limits show the upper- and lower quartile ranges, the center lines show the median values. The whiskers show the minimum and maximum values. h, A qq-plot of p-values from permutation analysis (n = 1,000 permutations) of variant burden in the SCHEMA consortium² data for upregulated genes in LRP1- and LDLR-KD neurons (n = 1,713). The p-values were calculated using Fisher’s exact test. The pink data point highlights enrichment of the gene set in question, the grey points demonstrate the permutation results.