Cell-type-specific transcriptional responses of leptomeninges from WT, Tlr4ECKO, and Tlr4MKO mice to neonatal E. coli infection.

(A) Schematic of the leptomeninges. (B) Immunostaining of P6 mouse leptomeninges viewed as a whole-mount (left) or as a transverse section through a series of Z-planes (right). The vertical black bars indicate the extent of the leptomeninges. Scale bars, 20 µm. (C) UMAP plot of leptomeninges snRNAseq from WT, Tlr4ECKO, and Tlr4MKO mice at P6, from both uninfected control and with E. coli infection, with cell clusters identified (see Figure 1 – figure supplement 1A-C). N=126,235 nuclei. Fb, fibroblast. (D) UMAP comparison of P6 leptomeninges snRNAseq from control vs. 24-hour E. coli infected mice of the indicated genotypes. (E) snRNAeq from four leptomeninges cell clusters (Fibroblast-arachnoid1, Fibroblast-arachnoid2, Fibroblast-pia, and Arachnoid barrier cells) showing differentially expressed genes in control vs. E. coli infected mice of the indicated genotypes at P6. (F and G) As in panel (E) except for ECs (F) and macrophages (G). DEGs, differentially expressed genes. (H) As in panel (F) and (G) except DEGs were related to NFkB and TNF-alpha signaling. (I) Summary of transcriptional responses of leptomeninges from WT, Tlr4ECKO, and Tlr4MKO mice to E. coli infection. (J) Summary of the effects of TLR4 signaling in leptomeningeal ECs.

Endothelial TLR4 drives ICAM-1 induction, increased vascular permeability, and macrophage activation during neonatal E. coli infection.

(A) Whole-mount leptomeninges from P6 WT, Tlr4ECKO, and Tlr4MKO mice, control or E. coli infected, immunostained for Cldn5 and ICAM-1 (upper panels) or immunostained for Cldn5 and incubated with fluorescent streptavidin (lower panels) to visualize the fate of intravascular sulfo-NHS-biotin. Scale bars: 20 µm. (B) Quantification of ICAM1 immunostaining relative to WT controls, as shown in (A). (C) Quantification of extravascular streptavidin binding relative to WT controls, as shown in (A). (D) Whole-mount leptomeninges from P6 WT and Tlr4ECKO mice, control or E. coli infected, immunostained for Cldn5 and either ASC or CD206 to visualize macrophages. Scale bars: 20 µm. (E) Quantification of ASC+ area (upper) and CD206+ area (lower), as shown in (D). Box plots show the median, interquartile range, and all individual data points. Each data point represents one image from a single mouse, with two locations imaged per mouse. Statistical comparisons (p-values) were calculated with the Wilcoxon rank-sum test. ns, non-significant (p > 0.05).

Claudin-5 redistribution in response to E.coli in leptomeningeal endothelial cells and in bEnd.3 cells.

(A) Whole-mount leptomeninges from P6 WT, Tlr4ECKO, and Tlr4MKO mice, control or E. coli infected, immunostained for Cldn5. Scale bars, 20 µm. (B) Quantification of Cldn5+ area in the leptomeninges, as shown in (A). (C) Control bEnd.3 cells and bEnd.3 cells exposed to live E. coli or live group B Streptococcus (GBS) for 1–6 hours were immunostained for Cldn5 and NF-κB. Scale bars, 20 µm. (D) Control bEnd.3 cells and bEnd.3 cells exposed to heat-killed E. coli or heat-killed group B Streptococcus (GBS) for 1–6 hours were immunostained for Cldn5 and NF-κB. Scale bars, 20 µm. (E) Quantification of experiments shown in (C) and (D). First plot, quantification of log10-transformed cytoplasmic-to-plasma membrane (Cyto/PM) intensity ratio for Cldn5 in bEnd.3 cells exposed to live E. coli for the indicated time in hours. Second plot, quantification of log10-transformed nuclear-to-cytoplasmic (Nuc/Cyto) intensity ratio for NF-κB in bEnd.3 cells exposed to live GBS for the indicated time in hours. The third and fourth plots are analogous to the first and second plots, except that bEnd.3 cells were exposed to heat-killed (HK) E. coli or GBS. Box plot in (B) shows median, interquartile range, and all individual data points. Each in vivo data point represents one image from a single mouse, with two locations imaged per mouse. For plots in (E), each data point represents a single bEnd.3 cell analyzed from representative images in three biological replicates. Statistical comparisons (p-values) were calculated with the Wilcoxon rank-sum test. ns, non-significant (p > 0.05).

TLR4 signaling controls NF-κB activation, tight junction dynamics, and barrier integrity in bEnd.3 cells exposed to E. coli.

(A) WT and Tlr4KO bEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 1 hour or 4 hours, were fixed and immunostained for Cldn5 and NF-κB and stained with DAPI. Scale bars: 20 µm. (B) WT and Tlr4KO bEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 1 hour or 4 hours, were fixed and immunostained for Cldn5 and ZO-1 and stained with CellTrace. Scale bars: 20 µm. (C) Quantification of log10-transformed nuclear-to-cytoplasmic (Nuc/Cyto) NF-κB intensity ratio in WT and Tlr4KO bEnd.3 cells, with or without exposure to E. coli, as shown in (A). (D) Quantification of log10-transformed cytoplasmic-to-plasma membrane (Cyto/PM) Cldn5 intensity ratio in WT and Tlr4KObEnd.3 cells, with or without exposure to E. coli, with ZO-1 marking the plasma membrane, as shown in (B). (E) WT and Tlr4KO bEnd.3 cells were grown to confluence on biotinylated gelatin-coated coverslips, were or were not exposed to E. coli for 4 hours, were then incubated with fluorescent streptavidin, and were finally fixed and immunostained for ZO-1. Representative images are shown with merged (upper) and streptavidin-only (lower) channels. Scale bars: 20 µm. (F) Quantification of streptavidin+ area in WT and Tlr4KO bEnd.3 cells with or without exposure to E. coli, as shown in (E). Wells coated with gelatin alone or biotinylated gelatin (without cultured cells) served as negative and positive controls, respectively. Box plots show median, interquartile range, and all individual data points. Each data point represents a single cell (C and D) or one image field (F) from representative images in three biological replicates. Statistical comparisons (p-values) were calculated with the Wilcoxon rank-sum test.

Comparisons of Cldn5 localization with junctional, plasma membrane, and trafficking markers in WT and Tlr4KObEnd.3 cells with or without E. coli exposure.

(A-C) WT and Tlr4KO bEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 4 h, were fixed and immunostained for Cldn5 and the indicated markers. Scale bars: 20 µm. (D) Quantification of overlap between Cldn5 and β-catenin, GLUT1, PECAM1, and ZO-1, with each data point representing a 100 µM x 100 µM region of interest (ROI). (E) WT and Tlr4KO bEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 4 h, were fixed and immunostained for Cldn5 and the indicated markers. Scale bars: 20 µm. (F) Quantification of overlap between Cldn5 and EEA1, LAMP2, and Rab7, as in (D). (G) WT and Tlr4KO bEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 4 h, were fixed and immunostained for Cldn5 and the indicated markers. Scale bars: 20 µm. (H) Quantification of overlap between Cldn5 and Rab11, PDI, and RCAS1, as in (D). Box plots show median, interquartile range, and all individual data points. Each data point represents one image from representative images in three biological replicates. Statistical comparisons (p-values) were calculated with the Wilcoxon rank-sum test.

In bEnd.3 cells, TLR4 signaling plays a central role in the transcriptional response to E. coli.

(A) Volcano plot of differential transcript abundances in WT bEnd.3 cells, with or without exposure to E. coli. The plot shows transcript-level log2 fold change (LFC) on the horizontal axis versus −log10(FDR) on the vertical axis. Red points mark significantly changed transcripts (FDR < 0.05 and |LFC| ≥ 1); grey points are not significant. (B) Volcano plot of differential transcript abundances in Tlr4KO bEnd.3 cells, with or without exposure to E. coli, plotted as in (A). Note the change of scale of the vertical and horizontal axes between (A) and (B). (C) Pathway effect map from ssGSEA Hallmark scores. Each symbol is a pathway. The horizontal axis is the effect of E. coli exposure on WT bEnd.3 cells (ΔssGSEA = E. coli exposed – not exposed) and the vertical axis is the effect of E. coli exposure on Tlr4KO bEnd.3 cells. Symbol size encodes the within-genotype significance [max(−log10 FDR) across both genotypes], and the color code indicates the genotype × E. Coli exposure interaction FDR. The 45-degree diagonal line represents equal effects for the two genotypes. (D–F) ssGSEA pathway activity based on genotype and E. coli exposure for three Hallmark immune response pathways: TNFα signaling via NF-κB (D), Inflammatory response (E), and IL6–JAK–STAT signaling (F). Symbols represent ssGSEA scores for individual samples. The vertical axis shows the interaction FDR score from a linear model for genotype x E. coli exposure. (G) Transcript abundance changes for 12 immune system genes in WT and Tlr4KO bEnd.3 cells, with or without exposure to E. coli. (H) Heatmap of transcript abundance changes in WT and Tlr4KObEnd.3 cells, with or without exposure to E. coli, for TNFα signaling via the NF-κB pathway. Rows show the transcripts with the greatest changes; values are variance-stabilized transformation (VST) z-scores derived from DESeq2-normalized counts. Columns are arranged by genotype and condition. (I) Scatterplot showing transcript abundance changes in WT and Tlr4KO bEnd.3 cells, with or without exposure to E. coli, for TNFα signaling via the NF-κB pathway. Each symbol represents one gene. Red points indicate genes significant for either genotype (FDR < 0.05). The top 15 genes by effect size are labeled. Note the different scales for the horizontal and vertical axes. The diagonal line represents equal effects for the two genotypes.

Specificity of VECad-CreER and Lyz2-Cre in leptomeninges whole-mounts.

(A) Lyz2Cre-mediated recombination of the loxP-stop-loxP (LSL) reporter ROSA26-LSL-tdT-2A-nlsGFP shows tdTomato expression in a subset of CD206+ immune cells. (B) Cdh5Cre-mediated recombination of the loxP-stop-loxP (LSL) reporter ROSA26-LSL-tdT-2A-nlsGFP shows tdTomato expression in all ECs, visualized with Cldn5 immunostaining. Scale bars, 20 µm.

snRNAseq cell cluster assignments and macrophage transcriptome responses to infection.

(A) UMAP plot of leptomeninges snRNAseq data from WT, Tlr4ECKO, and Tlr4MKO mice at P6, either uninfected control or with E. coli infection, with cell clusters identified. N=126,235 nuclei. Fb, fibroblast. (B) snRNAseq transcript abundances among the six major leptomeninges cell clusters for a set of 33 genes, for which the transcript abundances distinguish these clusters. We note that cluster Fb-arachnoid2 was originally mis-assigned as a dura fibroblast cluster (“fibroblast dura3”) in Wang et al. (2023). (C) UMAP plots of leptomenenges snRNAseq showing the abundances of transcripts that are highly enriched in each of the six major leptomeninges cell clusters. (D) snRNAeq for the macrophage cell cluster showing abundant and differentially expressed genes in control vs. E. coli infected mice of the indicated genotypes at P6.

Genes regulated by E. coli infection in leptomeningeal endothelial cells and macrophages.

(A) Genes regulated by E. coli infection in leptomeningeal endothelial cells. Left panel, most differentially regulated genes based on adjusted p-value. Right two panels, most differentially regulated genes based on fold-change. (B) Genes regulated by E. coli infection in leptomeningeal macrophages. Left panel, most differentially regulated genes based on adjusted p-value. Right two panels, most differentially regulated genes based on fold-change.

Infection responses in the leptomeninges and adjacent cerebral cortex in WT and Tlr4ECKO mice.

(A) UMAP plot of leptomeninges snRNAseq data from WT and Tlr4ECKO mice at P6, either uninfected control or with E. coli infection, showing Icam1 transcripts. (B) Whole-mount cortical surface, with attached leptomeninges, from P6 WT and Tlr4ECKO mice, control or E. coli infected, immunostained for Cldn5 and ICAM1. Upper panels, stacked Z-planes at the level of the subarachnoid space; lower panels, stacked Z-planes at the level of the adjacent cerebral cortex. Scale bars, 20 µm. (C) Quantification of ICAM1 immunostaining in the subarachnoid space and adjacent cerebral cortex, as shown in (B). Note the different scales for the vertical axis in the two plots. (D) As in (B), except immunostained for Cldn5 and ASC. The paired images in the upper and lower rows are part of the same confocal imaging file. The upper row of images here reproduces the upper row of images in Figure 2D. Here, each image in the lower row shows the same X-Y area as shown in the image above it, but at a greater Z-depth to visualize the underlying cerebral cortex. (E) Quantification of ASC immunostaining in the cerebral cortex, as shown in the lower row of images in (D). (F) As in (B), except immunostained for Cldn5 and CD206. The paired images in the upper and lower rows are part of the same confocal imaging file. The upper row of images reproduces the lower row of images in Figure 2D. Here, each image in the lower row of images shows the same X-Y area as shown in the image above it, but at a greater Z-depth to visualize the underlying cerebral cortex. These data are not quantified because CD206 cells were so rare in the cortex in all samples that accurate quantification would have been difficult.

ZO1 in leptomeningeal endothelial cells with or without E. coli infection.

(A) Whole-mount leptomeninges from P6 WT and Tlr4ECKO mice, control or E. coli infected, immunostained for Cldn5 and ZO1. Scale bars, 20 µm. (B) Quantification of ZO1 immunostaining in the subarachnoid space, as shown in (A).

CRISPR KO of Tlr4 in bEnd.3 cells

(A) Functional testing of three bEnd.3 clones following CRISPR KO of Tlr4. For each clone, the TLR4-mediated NFkB response to E. coli exposure (migration of NFkB from cytoplasm to nucleus) was quantified by immunostaining, as shown in Figure 3C-E. (B) Sanger sequencing from two plasmid subclones carrying genomic PCR products encompassing the region targeted by the Tlr4 guide RNA from bEnd.3-Tlr4KO-m1 cells. Each sequencing run shows a distinctive frameshift mutation. The sequences show the WT sequence on the left and the mutant sequence (with the altered nucleotides in red) on the right. Left, a single nucleotide insertion. Right, a two-nucleotide deletion. (C) Compendium of ten plasmid subclone sequences from each of the three bEnd.3 clones shown in (A). Each clones shows three distinct alleles, all of which differ from WT and all of which produce a frameshift. No WT sequences were observed in any subclone. These data imply the Tlr4 gene is present at three copies in bEnd.3 cells, and that each subclone is likely null for Tlr4 function.

The streptavidin leak assay requires a confluent monolayer of bEnd.3 cells.

(A) Among bEnd.3 cells, junctional localization of Cldn5 requires cell-cell contact as shown by immunostaining of cells grown at different densities. (B) bEnd.3 cells were grown at either low or high density on biotinylated gelatin-coated coverslips, incubated with fluorescent streptavidin, and then fixed and immunostained for ZO-1. In regions where the bEnd.3 cells had not formed a contiguous network of tight junctions, the streptavidin gained access to the underlying biotinylated gelatin.

Cldn5 dynamics in bEnd.3 cells following exposure to E. coli.

(A) The overlap between Cldn5 and β-catenin, GLUT1, PECAM1, and ZO-1, in WT and Tlr4KObEnd.3 cells, either not exposed to E. coli (control) or exposed to E. coli for 4 hours. Shown here is a more granular quantification of the experiment shown in Figure 5A-D, with each datapoint representing a 100 µM x 100 µM region of interest (ROI). (B) bEnd.3 cells were pre-incubated for 30 minutes with the indicated fluorescent tracer, then either not exposed to E. coli (control) or exposed to E. coli for 4 hours, and finally fixed and immunostained for Cldn5. Scale bars, 20 µm. (C) Quantification of % overlap between tracer and Cldn5 from the experiment shown in (B). (D) Cldn5 internalization and recovery during 1 hour of E. coli exposure, followed by washing away of the E. coli, and then an additional 3 hours of incubation in medium without bacteria. (E) Quantification of the experiment shown in (D). The metric on the y-axis is (cytoplasmic signal – plasma membrane signal)/(cytoplasmic signal + plasma membrane signal). 100% cytoplasmic localization corresponds to 1.0. 100% plasma membrane localization corresponds to −1.0. 50% cytoplasmic localization + 50% plasma membrane localization corresponds to 0.0. Box plots show median, interquartile range, and all individual data points. Each data point represents one image from representative images in three biological replicates. Statistical comparisons (p-values) were calculated with the Wilcoxon rank-sum test.

Transcriptome analysis of WT and Tlr4KO bEnd.3 cells, with or without exposure to E. coli.

(A) Principal component analysis showing the relatedness of the 14 bEnd.3 RNAseq samples. Two of the four orange symbols are nearly superimposed. (B) Pearson correlations for the 14 bEnd.3 RNAseq samples. (C) Pathway-level effects, showing the difference between WT and Tlr4KO bEnd.3 cells for control vs. E. coli exposure. Red, false discovery rate (FDR) less than 0.01. Orange, FDR between 0.01 and 0.05. Green, FDR between 0.05 and 0.25. Grey, FDR greater than 0.25. (D-F) Individual transcript abundances in the 14 bEnd.3 RNAseq samples for three GSEA categories. The transcriptome responses to infection are minimal in Tlr4KO bEnd.3 cells.